One problem with the split-half method is that the reliability estimate obtained using any random split of the items is likely to differ from that obtained using another. One solution to this problem is to compute the Spearman-Brown corrected split-half reliability coefficient for every one of the possible split-halves and then find the mean of those coefficients. This is the motivation for Cronbach’s alpha.

Cronbach’s alpha is superior to Kuder and Richardson Formula 20 since it can be used with continuous and non-dichotomous data. In particular, it can be used for testing with partial credit and for questionnaires using a Likert scale.

**Definition 1**: Given variable *x*_{1}*, …, x _{k}* and

*x*

_{0}= and

**Cronbach’s alpha**is defined to be

**Property 1**: Let *x _{j} = t_{j} + e_{j}* where each

*e*is independent of

_{j}*t*and all the

_{j}*e*are independent of each other. Also let

_{j}*x*

_{0}= and

*t*

_{0}= . Then the reliability of

*x*

_{0}≥

*α*where

*α*is Cronbach’s alpha.

Here we view the *x _{j}* as the measured values, the

*t*as the true values and the

_{j}*e*as the measurement error values. Click here for a proof of Property 1.

_{j}**Observation**: Cronbach’s alpha provides a useful lower bound on reliability (as seen in Property 1). Cronbach’s alpha will generally increase when the correlations between the items increase. For this reason the coefficient measures the internal consistency of the test. Its maximum value is 1, and usually its minimum is 0, although it can be negative (see below).

A commonly-accepted rule of thumb is that an alpha of 0.7 (some say 0.6) indicates acceptable reliability and 0.8 or higher indicates good reliability. Very high reliability (0.95 or higher) is not necessarily desirable, as this indicates that the items may be entirely redundant. These are only guidelines and the actual value of Cronbach’s alpha will depend on many things. E.g. as the number of items increases, Cronbach’s alpha tends to increase too even without any increase in internal consistency.

The goal in designing a reliable instrument is for scores on similar items to be related (internally consistent), but for each to contribute some unique information as well.

**Observation**: There are an number reasons why Cronbach’s alpha could be low or even negative even for a perfectly valid test. Two such reasons are reverse coding and multiple factors.

**Reverse coding**: Suppose you use a Likert scale of 1 to 7 with 1 meaning strongly disagree and 7 meaning strongly agree. Suppose two of your questions are: Q1: “I like pizza” and Q20: “I dislike pizza”. These questions ask the same thing, but with reverse wording. In order to apply Cronbach’s alpha properly you need to reverse the scoring of any negatively phrased question, Q20 in our example. Thus if a response to Q20 is say 2, it needs to be scored as 6 instead of 2 (i.e. 8 minus the recorded score).

**Multiple factors**: Cronbach’s alpha is useful where all the questions are testing more or less the same thing, called a “factor”. If there are multiple factors then you need to determine which questions are testing which factors. If say there are 3 factors (e.g. *happiness with your job, happiness with your marriage* and *happiness with yourself*), then you need to split the questionnaire/test into three tests, one containing the questions testing factor 1, one with the questions testing factor 2 and the third with questions testing factor 3. You then calculate Cronbach’s alpha for each of the three tests. The process of determining these “hidden” factors and splitting the test by factor is called Factor Analysis (see Factor Analysis).

**Example 1**: Calculate Cronbach’s alpha for the data in Example 1 of Kuder and Richardson Formula 20 (repeated in Figure 1 below).

**Figure 1 – Cronbach’s Alpha for Example 1**

The worksheet in Figure 1 is very similar to the worksheet in Figure 1 of Kuder and Richardson Formula 20. Row 17 contains the variance for each of the questions. E.g. the variance for question 1 (cell B17) is calculated by the formula =VARP(B4:B15). Other key formulas used to calculate Cronbach’s alpha in Figure 1 are described in Figure 2.

**Figure 2 – ****Key formulas for the worksheet in Figure 1**

Since the questions only have two answers, Cronbach’s alpha .73082 We see that this is the same as the We see that this is the same as the KR20 reliability calculated for Example 1 of Kuder and Richardson Formula 20.

**Observation**: If the variances of the *x _{j}* vary widely, the

*x*can be standardized to obtain a standard deviation of 1 prior to calculating Cronbach’s alpha.

_{j}**Observation**: To determine how each question on a test impacts the reliability, Cronbach’s alpha can be calculated after deleting the *i*th variable, for each *i ≤ k.* Thus for a test with *k* questions, each with score *x _{j}*, Cronbach’s alpha is calculated for for all

*i*where = .

If the reliability coefficient increases after an item is deleted, you can assume that the item is not highly correlated with the other items. Conversely, if the reliability coefficient decreases, you can assume that the item is highly correlated with the other items.

**Example 2**: Calculate Cronbach’s alpha for the survey in Example 1, where any one question is removed.

The necessary calculations are displayed in Figure 3.

**Figure 3 – Cronbach’s Alpha for Example 2**

Each of the columns B through L represents the test with one question removed. Column B corresponds to question #1, column C corresponds to question #2, etc. Figure 4 displays the formulas corresponding to question #1 (i.e. column B); the formulas for the other questions are similar. Some of the references are to cells shown in Figure 2.

**Figure 4 – ****Key formulas for worksheet in Figure 3**

As can be seen from Figure 3, the omission of any single question doesn’t change the Cronbach’s alpha very much. Removal of Q8 affects the result the most.

**Observation**: Another way to calculate Cronbach’s alpha is to use the **Two Factor ANOVA without Replication** data analysis tool on the raw data and note that:

**Example 3**: Calculate the Cronbach’s alpha for Example 1 using ANOVA.

We begin by running Excel’s **Anova: Two Factor without Replication** data analysis tool using the data in range B4:L15 of the worksheet shown in Figure 1.

**Figure 5 – Calculation of Cronbach’s alpha using ANOVA**

As you can see from Figure 5, Cronbach’s alpha is .73802, the same value calculated in Figure 1.

**Observation**: Alternatively, we could use the Real Statistics **Two Factor ANOVA** data analysis tool, setting the **Number of Rows per Sample** to 1. We can also obtain the same result using the following supplemental function.

**Real Statistics Functions**: The following functions are provided in the Real Statistics Resource Pack:

**CRONALPHA**(R1, *k*) = Cronbach’s alpha for the data in range R1 if *k* = 0 (default) and Cronbach’s alpha with *k*th item (i.e. column) removed if *k* > 0

**CALPHA**(R1): array function which returns a row of Cronbach’s alpha for R1 with each item removed

Thus for the data in Example 1, we can obtain the results shown in Figure 1 and 3 using CRONALPHA(B4:L15) = .738019 and CRONALPHA(B4:L15, 1) = .76321. Also the formula CALPHA(B4:L15) can be used to produce the results shown in range B43:L43 of Figure 3.

**Example 4**: Calculate Cronbach’s alpha for a 10 question questionnaire with Likert scores between 1 and 7 based on the 15 person sample shown in Figure 6.

**Figure 6 – Calculation of Cronbach’s alpha for Example 4**

As you can see from Figure 6, Cronbach’s alpha is 0.59172, a little below the generally acceptable range. We get the same answer by using the supplemental formula in the Real Statistics Resource Pack, namely CRONALPHA(B4:K18) = 0.59172.

**Real Statistics Data Analysis Tool**: The Real Statistics Resource Pack provides the **Reliability** data analysis tool which can be used to calculate Cronbach’s alpha.

We now show how to calculate Cronbach’s alpha for Example 4 using the **Reliability** data analysis tool. Start, as usual, by pressing **Ctrl-m** and choose the **Reliability** option from the menu that appears. Fill in the dialog box that appears as shown in Figure 7.

**Figure 7 – Dialog box for Reliability data analysis tool**

The output is shown in Figure 8. Cronbach’s alpha is shown in cell M3, while the Cronbach’s alpha values with one question removed are shown in range M8:V8, which is the same as the output from =CALPHA(B4:K18). Note too that the split-half measures are also shown in N12 and N13.

**Figure 8 – Cronbach’s alpha option of Reliability data analysis tool**

Additional information about Cronbach’s Alpha can be found by clicking here. This includes how to perform hypothesis testing on Cronbach’s Alpha, confidence intervals, statistical power and sample size requirements.

May I ask, can we run cronbach alpha test before having data in hand (e.g. before pilot study)?

You can run Cronbach’s alpha using the data from a pilot study.

Charles

Please I want to know if i can check the reliability of a “writing test” using this technique or it is just used for questionnaires!!!

How will you score the writing test?

Charles

pls am doing my project in which I want how to analysis it am having 20 questionnaire and it contains 17 questions each and respondent has to choose from strongly Agree or Agree or strongly disagree or disagree in their responses. how I calculate it using alpha

Abimbola,

You can calculate Cronbach’s alpha with Likert scale items as shown in Example 4 of the referenced webpage. Alternatively you can use the Real Statistics capabilities described on that webpage.

Charles

Can we use Cronbach’s Alpha to see if there is an agreement in the ratings (of features of some products) of two groups (gender groups or age groups for example)? If yes, how much sample size is enough for 6 features (rating 1 to 5).

Learner,

Cronbach’s alpha is not used for this purpose. To determine whether there is agreement in rating, you could use Cohen’s kappa, Fleiss’s kappa, ICC, Kendall’s W, etc., depending on the specifics of what you are trying to do. I suggest that you look at the website for more details about each of these approaches and see which one best fits with your needs. See the webpage

Reliability

Charles

Many thanks Professor. Earlier I found use of kappa coefficient and I’ve tried different measures like Cohen’s kappa, Fleiss’s kappa and Krippendorff’s Alpha, Kendall’s tau etc. using R, all gave me similar results regarding agreement in the ratings (of features of some products) of two groups (gender groups or age groups for example).

But I was wondering, what is the fundamental difference (difference between Cohen’s kappa, Fleiss’s kappa is clear to me). How to choose between these coefficient?

I found your way of explaining things is the best.

Learner,

A key difference relates to the type of input data. E.g. Cohen’s kappa supports two raters with categorical ratings, Fleiss’s kappa supports two or more raters with categorical ratings, Weighted kappa supports two raters with ordered ratings, ICC supports any number of raters with interval/continuous ratings.

Charles

Thanks a ton Professor. So simple explanation!!

What I could understood is that ICC is used to measure internal consistency of the raters (within each group) and the measures mentioned by you (Cohen’s kappa, Fleiss’s kappa and weighted kappa) measures agreement between

two or more group of raters. Please correct, if I am wrong.

Learner,

Sorry, but I don’t understand what you wrote.

Charles

I have spent all day reading psychometric testing books and this is the most helpful thing I’ve read on Cronbach’s alpha. Thank you so much – very clear.

Hello Charles

could you explain to me how I create a table with 5 separately calculated alpha-coefficients in SPSS?

Thank you

Sorry Alina, but I don’t use SPSS.

Charles

Hello,

I know that if you have several dimensions, you calculate Cronbach alpha for each dimension. But can you calculate it for the whole test? In case you want to know reliability for the test as whole. Or do you rather conclude on the whole reliability based on reliability of each dimension?

Thank you very much

If there are several dimensions you should use the Cronbach’s alpha for each dimension. Cronbach’s for the whole test won’t be that meaningful. If you need one value, you could use the smallest of the Cronbach’s alpha values, but it is better to simply use Cronbach’s for each dimension.

Charles

I have two groups of questions; First is Functional form of question (Positive side) , for example: If a door is closing how do you feel. One answer be selected by the user 1. I like 2. It must be that way, 3. I am Neutral, 4. I can live that way, 5. I dislike it that way

Second question: Dysfunctional (negative): If a door is not closing how do you feel. The answer be selected by the user 1. I like 2. It must be that way, 3. I am Neutral, 4. I can live that way, 5. I dislike it that way. How to can I certainly use Cronbach’s alpha

Shri,

If some of the questions reverse the scale then you should reverse the coding (i.e. 1 becomes 5, 2 becomes 4, 4 becomes 2 and 5 becomes 1), but I can’t tell from your example whether this is really a case of reverse coding or not.

Charles

I would like to know the answer for this question. Anyone who knows answer this.

Question: If you are provided with questionnaire consisting of 20 questions and you have the response in the form of a likert scale value(1 to 5) obtained form 25 respondents, then how will you find the factor loadings value?

If you are looking for factor loading values, you are probably referring to Factor Analysis, See the following webpage:

Factor Analysis

Charles

Hi Charles,

I want to know the minimum sample size required to safely calculate the test reliability (cronbach alpha) in a 4 choice mcq test consisting of 3 sections with 20 questions in each section. Each section measures different skills.

Also, kindly explain the method so I can determine the same for different variants of the test.

Regards

Hi Sarthak,

You need to calculate three Cronbach alphas, one for each section.

Regarding the sample size required, see the following webpage

http://www.real-statistics.com/reliability/cronbachs-alpha/cronbachs-alpha-continued/

Charles

I am new to this site/blog but it is exciting.

I am involved in HIV research at operation level. So at time we wonder how to measure internal consistence. For example we have this question:

In the last 12 months, have you experienced any of the following feelings because you are living with or affected by HIV? (Tick as many boxes as apply ) (read out)

1. I feel ashamed

2. I feel guilty

3. I blame others

4. I blame my self

5. I have low self esteem

6. I feel i should be punished etc.

All these are items that measure what is termed as Internal HIV stigma.

Kindly help guide me how to measure the internal consistency. This will guide whether I should drop some items.

Essentially you have 6 questions with a response of ticked or not ticked. You can assign the code 1 for the ticked items and 0 for non ticked items. This is the coding assuming that all the ticked items represent Internal HIV stigma; if some of the items are reverse coded (i.e. not ticked represents Internal HIV stigma) then simply interchange the roles of 0 and 1 for those items. You then calculate Cronbach’s alpha as described on the referenced webpage.

Charles

Hi charles,

im doing an experiment about voice perception.. i use likert scale in in 5 question. There are five voices. The questions are do u find the voice trusthworthy dominant aggressive etc. Can this method help me solve this?pls. A student from philippines

It depends on what you mean by “Can this method help me solve this?”. What is it that you want to solve?

You can certainly use Cronbach’s alpha with a Likert scale.

Charles

Hi Charles,

I am questioning if I did my Cronbach’s Alpha test correct. I have one cell with a score of -7.113. The rest of the scores fall between 2.25 and 0.22.

Should/can there be negative numbers? And should I be concerned about my one cell that has a score of 7?

I realize this may be an impossible question for you to answer without my work (most questions are a Likert scale with a range of 1 – 7).

Thank you in advance.

Pat,

Yes, I would need to see your data to understand your question. You can send me an Excel file to my email address, as shown on Contact Us.

Charles

Hi Charles,

I am looking for some guidance – – I have administered a pre-/post- survey, using a 5 point likert scale. Response prediction was a one-tailed shift, and this was true. Sample size was only 32, and most respondents shifted from a ‘3’ to a ‘2’ or ‘1’, so not a huge shift, but one that was pretty uniform. There were about 6 that did not shift at all. Although the shift was evident, my paired sample ttest returns a very high p-value, indicating no statistical significance. I am wondering if I should use a different analysis approach?? little out of my depth…

Kate,

Can you send me the data so that I can understand the situation better? See Contact Us for my email address.

Charles

Hi Charles,

I’m using a likert scale 1 to 5, 15 questions as feedback about facebook and moodle. Same set of questions apply to both facebook and moodle given to the same students (21 students). What is the best and proper approach? Seems like I cannot use the 0/1 coding. Please advice.

Depending on what you are measuring, you might need to compute two Cronbach’s alphas, one for moodle and the other for facebook. You can use the Likert scale as input to Cronbach’s alpha.

Charles

Hi Charles,

If the response is 1/0 the formula is …=(k/(k-1)…

if the response is 1, 2, 3, 4, 5…what would be the formula?

please advice

Tobias,

Sorry, but I don’t know which formula you are referring to. The formula that I use for Cronbach’s alpha is the same for dichotomous values as for Likert values.

Charles

I designed a 7th grade science test with 30 multiple choice items as part of an experimental investigation. Each question has four alternatives. Part I consists of 10 questions about the scientific method; Part II consists of 15 questions on the structure and organization of organisms, and; Part III consists of 5 questions about taxonomy. My question is: When I calculate reliability, should I calculate three separate Cronbach alphas (one for each part), or only one overall Cronbach’s alpha for the entire test? Also, is using the 0/1 coding (incorrect/correct) the correct procedure?

I appreciate your input and guidance.

Marilyn,

Since it looks like you are testing three different concepts, you should calculate three separate Cronbach alphas. The 0/1 coding (incorrect/correct) is correct.

Charles

Hi!

I would like to know if I can conduct cronbachs alpha on my own questionnaire?

For example, my questionnaire would be like this:

1. Advertising through tv or radio creates brand awareness.

2.Internet marketing creates brand awareness.

Then this would be in the form of a likert scale. 5-strongly agree then 1-strongly disagree.

Thanks for the help!

Pam,

Sure, you can use Cronbach’s alpha to test such things. It is important that if you are testing multiple concepts that you create a Cronbach’s alpha score for each concept based on the questions that relate to that concept.

Charles

For example,

there would be concepts about advertising, personal selling, sales promotion, and public relations. But this would be in just one questionnaire. Should i conduct cronbachs alpha separately for each concept?

Thanks!

Yes, you should conduct a separate Cronbach’s alpha for each concept.

Charles

May I ask something else, please?

1. If someone has a pre-post design with 60 students and wants to compare the significance of the pre-post differences (per critical thinking skill let’s say), why would they choose to do the comparison by performing z-tests with students’ percentages per skill before and after the intervention? Couldn’t they have just estimated the pre/post-means and run a t-test or a non parametric equivalent?

2. If they already had their z-tests for comparing separately whether the number (percentages) of students at each level of each skill has increased or decreased significantly, why would they continue with contigency tables and chi-square for checking whether students transition from skill-level to skill-level (low…high..etc) for each skill separeately, is significant or not? I mean do you think that this info really adds to their argument about the efficacy of their intervention regarding the target-skills?

Thank you very much.

Maria,

1. If you are trying to compare before and after data, the usual approach is to use a paired t-test of a paired Wilcoxon signed ranks test. Both are described on the website.

2. I don’t completely understand this question.

Charles

Thank you very much, Charles.

Let me clarify my 2nd question: if someone does find a significant gain in the learning outcomes of their interst by comparing pre-post data, is there actually a reason for them to search with chi-square whether the transition of students from “low to middle” or from “middle to high” level responses is significant too? Do you think that this info can really strengthen the argument about the efficacy of the intervention which was already shown, or it is practically useless?

Maria,

I don’t know what chi-square test you are referring to in the question “is there actually a reason for them to search with chi-square whether the transition of students from “low to middle” or from “middle to high” level responses is significant too?”

Charles

First of all, you are doing a really great job here. Thank you very much.

My question is: Can someone use Cronbach’s alpha in order to come up with conclusions about the reliability of a questionnare that includes only open-ended questions? These open-ended questions are supposed to elicit critical thinking skills from the respondents and they are coded per skill as “insufficient”, “low”, “middle” and “high”.

Maria,

You can use Cronbach’s alpha as long as you have an objective way to code the responses as “insufficient”, “low”, “middle” and “high”, and you can assign numeric values to these responses (e.g. as 0, 1, 2, 3).

Charles

Thank you very much for your prompt, clear answer.

Dear Sir,

Thanks for the further explanations. What if every student answered every question correctly? That’ll result in 11 for all cells M4-M15, correct? And therefore, the variance in the totals would be 0 (cell B21). How would you interpret such a result with respect to the alpha, please?

Thank you,

Marc

Marc,

If everyone answers the question correctly, in some sense the reliability would be perfect, i.e. 1, but not particularly interesting. For many of the reliability measurements, the mathematical result in extreme situation, such as this one, is counter-intuitive or not very helpful. Often because the variance is zero.

Charles

Charles,

Thank you very much indeed. That really helps my discussion and is a good contra argument

Marc

Hi Charles , I have a question here…

I am doing single subject. Questionnaire (Scale 1- 4) was administered 15 times with 3 subjects on 15 days which meant

1. 5 time baseline

2. 5 time intervention

3. 5 time withdrawn

My question here is how do I calculate cronbachs’ alpha with 15 time different phase?

Tiffany,

You would typically run Cronbach’s alpha multiple times, once per each of the time periods. This is of course impossible if you only have one subject (although you also say that you have three subjects).

Charles

Sir I need your help

Can’t you help me about kuder richard formula 20

I need exact explenation about kuder richard formula 20

thanks.

Sorry, but I don’t have anything else to add to what I responded earlier.

Charles

Hi there,

I have a very basic question. I am doing exploratory research with 500+ participants in predicting risk and thriving behaviours. My problem is that some of my outcome variables have only two items, thus my Cronbach alphas are always .50-.65 range. I’ve been reading that it is not even acceptable to run alpha analysis when you only have two items. Any thoughts for or against this practice?

There really isn’t much point in running Cronbach’s alpha with only two items, but you can calculate a result. Although I haven’t really thought it through, you might try using Cohen’s kappa to see whether there is agreement among the participants between the two items.

Charles

The variables are all continuous , and as I understand Cohen’s kappa is for inter-rater agreement on categorical variables.

Good point.

Charles

Hi,

For my thesis I have made two sections first is a multiple choice questionnaire and second is a 5 point likert scale. May I use split half method for the questionnaire and cronbach alpha for the rating scale? Please help.

You can use Cronbach’s alpha for the Likert scale. You can use either Cronbach’s alpha or split half for multiple choice questions.

Charles

Help pls! We have rows of necessities such as food, etc. and column of value like very important, important, etc. the total is always 50. And we have computed the alpha which resulted to -2.something. What does it interpret?

50 because 50 is our total respondents. This is a research our adviser told us to use cronbach but we dont know how to interpret it on our study. Our study is about minimum wage earners

Kim,

I am sorry, but I don’t completely understand the situation that you are describing. Are you saying that the alpha value is a negative value less than -2? This is a very unlikely result (sic).

Charles

Thank you for this Charles.

Can I please ask you to clarify my problem with a reliability test that I carried out pre and post survey. The cronbach’s alpha values were in the range of 0.73 to 0.90 for the pilot tests on 10 responses, deeming the scale to be reliable. However, when I ran a reliability test post survey with 82 responses, the cronbach’s alpha’s ranged from 0.57 to 0.83. The cronbach alpha values had decreased for some of the constructs and since as per the literature, constructs with values below 0.70 are said to have poor reliability, I am not sure what kind of inference I can make out of this situation for my research. Also , at this stage I cannot delete any items off the questionnaire because it is post survey. Can you please suggest an explanation for the constructs with lower alpha values and how I can defend their reliability for my research.

Unfortunately, a pilot test of 10 responses is not so much, and so it is entirely possible that the actual Cronbach’s alpha is different from what you found; in fact if you set the significance level to 95% (alpha = .05), you should assume that one out of 20 times you could get surprised in this way.

The .70 value is only a guideline. I have read others saying that .60 is also acceptable. This sort of problem happens. The only thing that I can suggest is to report your findings as they are. Clearly some of the alpha values are higher than .70, which is good. For the ones that are much lower, you can explain what you might do in the future differently.

Charles

Thank you for this Charles.

Your website has been most helpful.

Dear Charles,

Thanks for the crystal clear explanation, it was very helpful.

I have one question that I’ve been struggling with lately.

Suppose I have 6 items (i.e. questions), that 100 participants (i.e. cases) answered on a continuous scrolling bar yielding valued between -350 and 350. I’d like to know whether these 6 items are “internally consistent” so I proceed to computing Chronbach’s alpha. However, should I, or should I not mean-center the 6 responses of each subject before computing the alpha?

The reason of my question is this: I noticed that subjects have very consistent spatial biases in the way the use my scale, so I tried mean-centering the scores by subject. As a result, the magnitude of alpha was strongly decreased.

This seems to derive from the fact that mean-centering by subject decreases the “total variance” component of the alpha formula, more than it does the (sum of the) within-item variance. However, intuitively, I don’t understand why differences in the way subjects use the scale (independently of what items the scale represent) should affect a measure of its internal reliability.

Surely, I’m getting something wrong, though I don’t quite know what.

Any ideas?

Thanks!

gabe

Gabe,

It depends on what you mean by “mean centering”. If you calculate the mean for each question and then for each participant and each question you subtract the mean for that question from the score of that participant for that question, Cronbach’s alpha will be the same.

Charles

Hi,

I created and performed an attitudinal survey regarding how people believe that are being treated in a particular company. I am looking at factors correlated with and/or those that predict work-life balance. I have extracted a number of factors from the literature which are associated with work life satisfaction, and I have adapted questions accordingly. I have administered the survey and now am trying to make sure it is a reliable and valid indicator. I used your (awesome) tool and have a high Cronbach alpha (>.90). What type of test should I run for validity? It may be that I publish this, so I want to make sure I can at least comment on some psychometric properties. Is Cronbach enough and, if not, what would you recommend? I am not going for the most robust analysis in the world, just something that provides that this is a reasonably credible instrument. Thanks!

Dan

Dan,

Here is a response to a similar question that I found on another site. I hope that you find it useful.

Charles

Julia B. Smith · 41.59 · 90.86 · Oakland University

If this is too basic, I apologize, but…

Reliability examines whether you are getting consistent information. Validity examines whether the information you are getting measures what you think it measures. My basic example is a scale – I can get on the scale 15 times in a row, get the same information each time, and that is reliable. But if I want to use the scale to tell me how tall I am, that is not valid.

Consequently, using a statistical tool to examine validity is entirely anchored in what it is you think is being measured, and what other source of information you have to compare it to. If I have a survey that I think measures depression, I am going to want to compare the results on that survey to results on an instrument that is already known to measure depression. That statistical analysis would require correlation, but the basis of the validity argument rests on the other source of information (and what is known about it).

Recall that there are several types of validity (content validity, construct validity, outcome validity, etc.), and each requires a different kind of examination. I have attached an article that does a nice job of discussing the problem of testing validity applied to student evaluation of college instructors.

Onwuegbuzie_et_al2007.pdf

Hi,

Can i use cronbachs alpha for satisfaction questions on a scale 1 to 100, with 100 being the most satisfied?

Ben,

Yes, you can use Cronbach’s alpha in this cazse.

Charles

Dear Charles,

first of all: Thanks for this awesome website. It already helped a lot!

My question: I already made all the calculations to get to the point of Figure 3 (Example 2). Now I need to figure out, which of my items can be omitted. You wrote that the one with the lowest alpha value affects the result the most. So it’s better to remove items with the highest alpha values? In the next step the omission of one item changes the alpha value of the others, right? Does this influence order?

Thanks a lot!

Christian

Dear Christian,

I am pleased that you find the website helpful.

Which item to remove really depends on what your objective is. If you objective is to increase reliability then you would want to remove the one with the lowest alpha value.

Removal of one item will change the values of the others. I don’t know how it impacts the order, but I suspect that the order can change.

Charles

Dear Charles,

thank you for your response.

As I already have about 20 items and due to this a high alpha (~0,95) I want to remove several items.

That’s why I’m wondering, which items would be best to keep and which items I can remove from my list.

Christian

Excuse me Charles,

just to get things clear:

Would you, in my case, recommend to delete the items with the lowest alphas until I only have a few items (say about 5) left, but still have an acceptable overall alpha for the scale?

Christian

Christian,

It is hard for me to say since I don’t know your objective. If you would prefer an example with fewer questions, then I would probably choose the course of action that you have described, especially since your Cronbach’s Alpha is 95% (which is quite high), and so it seems that all the questions are pretty much testing the same thing.

Charles

Hello

require your assistance to calculate the Cronbach’s alpha for my questionnaire

it consists 4 Section and in each section there are a number of questions and the last section two open questions

it is measured as well

when I calculate the number of items

how I do it

I count the number of question or what

if you have an email I want to send the result for you please

regards

You can send an Excel spreadsheet with your data to my email address, which you can find at Contact Us.

Charles

Hi

My questionnaire was administered four times with 15 subjects on two days. How do I calculate cronbachs’ alpha? Should I take average of all four times for day 1 and day 2 and then average for the two days? Which values should I consider?

When you say that your “questionnaire was administered four times with 15 subjects on two days” does this mean

(1) that in total you have 15 subjects who answered the questionnaire, each subject answering the questionnaire one time, or

(2) as in (1) except that you had 15 subjects four times, and so you had 60 subjects in total answering the questionnaire, or

(3) 15 subjects answered the same questionnaire four times?

If the situation is (1) or (2), you can simply apply Cronbach’s alpha on all 15 (or 60) subjects, ignoring the day that they answered the questionnaire (unless you think that for some reason this is relevant).

If the situation is (3), then you should probably report four different Cronbach’s alpha values, one for each time the 15 subjects answered the questionnaire. I might give a different answer if I knew why they needed to answer the same questionnaire four times in two days.

Charles

They answered the same questionnaire four times on each of the two days. this was done to see any significant difference in subjective responses before and after a rest period.

so we had following experimental design:

day 1 (1) day 1 (2) day 1 (3) day 1 (4)

s1-s15 s1-s15 s1-s15 s1-s15

day 2 (1) day 2 (2) day 3 (3) day 4 (4)

s1-s15 s1-s15 s1-s15 s1-s15

You can use repeated measures ANOVA to determine whether there is a significant difference between the test results.

You can use Cronbach’s alpha to measure the reliability of the test, but in this case, I would calculate four separate Cronbach alphas.

Charles

Hi, I have problem with cronbach’s alpha. I only have 3 questions for one variable, in which the alpha is only 0.372. However, when I deleted one question, it is still low as 0.4978. Is it still possible to use the data for regression, while the alpha is only 0.4978? thanks

Dani,

I would think that you could still use regression in this case. But you should make sure that the model has a good fit.

Charles

hello !

I would like to do a Cronbach for my test but somehow i am not sure it is pertinent.

I’m not confident in doing this kind of work so i will explain detailfully.

Actually, there is a question and 7 possibilities of answer but only one is the good one.

So i wrote 1 if succeed, and 0 if didn’t succeed.

Then i did the cronbach test. 28 questions. 11 persons.

So it means (k/(k-1)) = 28/27 = 1,04. Isn’t it ?

Then i should calculate the variance for each question, one by one

(so it means if i get only 1 for all the answer, variance will be 0). Then i add them.

And then , this is where i have difficulties : i should calculate the variance between the total of the the 28 questions ? Or variance between each person ?

If i do it between each person i get -0,77

If i do it between the questions i get 0,46

And if i fake to have almost always good answer, i get 1,03 for example. Is it normal ?

Also I tried to do it question per question to find out which of the questions put the level so down. Is there a special formula for this ? Or should i apply it 28 times without one item.

Thank you so very much for your answer !

Lilla,

I wasn’t able to follow all the steps you listed, but in any case you have the following options for computing Cronbach’s alpha

1. Use any of the techniques described on the referenced webpage

2. View the Excel worksheets that carry out the various techniques described on the referenced webpage. You can download these worksheets for free from the webpage Examples Workbooks.

3. Use the Real Statistics

CRONALPHAfunction or theCronbach’s Alphaoption on the Real StatisticsReliabilitydata analysis tool.Charles

hi,

i want to test the reliability of a multiple choice test. how do i code the responses and how do i find the reliability using cronbach alpha

Natasha,

The simplest way to code the response is 1 for a correct answer and 0 for an incorrect answer. You can then calculate Cronbach’s alpha using any of the approaches described on the referenced webpage.

Charles

Hi Charles,

Thank you for the description of Cronbach’s alfa, I really found it the best from several references, I reviewed in the past week.

I’m a PHD student conducted a questionnaire on 50 participants asking them 30 questions, each question has 3 multiple choices. Each choice support one of three leadership approaches from different perspectives (reordered randomly).

Because my thesis is only interested in the middle approach between two extremes; when I tried to calculate Cronbach’s alfa I considered the vote to the middle approach to be 1 and any vote to the other two approaches to be 0. Is this correct? The middle approach got 89% of the total votes (1500). I was expecting alfa to be very high because of that but when I calculated alfa it was very low (0.164). Please advice.

Many thanks.

Sherif,

The approach of 1 for the “correct” answer and 0 for an incorrect answer is the usual approach. I can’t say why the alpha value was so low, but usually this a sign of a problem showing low value for many call internal consistency, possibly because the test is measuring more than one thing (“factor”).

Care must also be taken with reverse coded questions (as described on the referenced website), although with 89% correct answers perhaps this is a less likely cause for concern in your case.

Another possible reason for a low alpha value is that the assumption for using this measure is violated (namely what is called “tau-equivalence”). Cronbach’s alpha is commonly used despite violations of this assumption. For this and many other reasons a lot of people don’t believe that Cronbach’s alpha is really a measure of reliability, despite the fact that it is commonly used and insisted on by many journals.

The fact that 89% of all the responses were correct seems very high. I can’t say whether this is contributing to the low alpha or not.

Charles

You are my lifesaver!

I have trouble with calculate cronbach alpha until I saw this page.

I finally got my cronbach alpha answer through Excel which is written by you.

Your posting is really helpful to me.

Thank you!

Good to hear that the information was helpful.

Charles

Can I find the reliability of a test using excel?

Kristine,

One approach may be to use Cronbach’s alpha as described on the referenced webpage.

Charles

Hi! I want to know the reliability of a 4o-item Math Test. How would I do that in SPSS? What data am I going to enter?

Kristine,

Sorry, but I am not familiar with SPSS. The Real Statistics Resource Pack provides such capabilities.

Charles

Dear Charles

This is a wonderful website.

Kindly assist I am administering a questionnaire in a census survey of 86 target respondents. The questionnaire has a total of 23 questions some of which have sub items. How many sample respondents should I use for calculating cronbach alpha in the pre-test?

Joseph

Joseph,

You can use the Statistical Power and Sample Size data analysis tool. Choose the Cronbach’s Alpha and Sample Size options.

Charles

Hi! How do you conduct cronbach alpha when the variables are income, consumption,a likert scale, and a dummy variable?

Justin,

Are using Cronbach’s alpha for a questionnaire? What do you mean by there are two variables? Do you mean two questions or two types of questions: those relating to income and those relating to consumption?

Charles

Yes, there are separate questions for income and consumption and we also have a question where it is answerable by 1 or 0 (dummy variable). We dont know how to do the cronbachs alpha for that dummy variable.

Justin,

Cronbach’s alpha can be used even when the questions take a value 1 (say for correct) and 0 (say for incorrect). This would cover True/False and multiple choice questions. The problem is that you can’t use Cronbach’s alpha with only one question. After all, Cronbach’s alpha measures internal consistency, but withonly one question there is no internal consistency to measure. The same is true for Kuder and Richardson.

Charles

Hi Charles,

I would like to know what kinds of questions could increase my Cronbach alpha? I did some readings to increase the quality of my questions in my questionnaire and after some piloting I have added 5 more items to the questionnaire but my alpha is still below 0.7 and I should delete some which are really critical questions. Do you have any suggestions? I feel so desperate.

Best,

Negar

Negar,

It is possible that your questionnaire is measuring more than one thing. In such cases, Cronbach’s alpha will be reduced. Instead of deleting questions, for the purposes of computing Cronbach’s alpha, you may need to split the questionnaire into two (or more) parts, one part for each group of questions that are measuring the same thing. In this way, you might be able to bring Cronbach’s alpha up to 0.7 on each part of the questionnaire.

If necessary, factor analysis can be used to figure out how to do the splitting of the questionnaire.

Charles

thnx for your help Charles

Hello Charles

I’m phd student and using questionnaire in my thesis .

I have 4 variable as independent and one dependent variable , each variable had 7-6 question. I tested Cronbach’s alpha for each variable in my pilot and remove and delete 3 to 4 question to increase my Cronbach’s alpha to .8 or .9 ,

and continue my data collection with 4-5 question for each variable.

now

i should analyses my main data with PLS,

but now, my Cronbach’s alpha decrease a gain and are less than .7 for each variable, and now i can not delete some question again.

what should i do now?

i really confuse.

thank

Nana

Is PLS partial least squares regression? Are the revised values for Cronbach’s alpha based on the latent variables (after applying PLS)?

If this is the situation, then I am not sure what the best approach is. If Cronbach’s alpha is just slightly less than .7 then I would worry about it too much.

Charles

thank you for your response. I have one more question. Can I do ANOVA by three schools? Is it possible to do ANOVA when you have ordinal data?

I don’t exactly know what you mean by doing ANOVA by three schools. You can do ANOVA with ordinal data, but I would not recommend it. You should use a non-parametric test instead (e.g. Kruskal-Wallis).

Charles

Excuse me sir. can I use average score of each teacher.

It depends on what you are measuring. If each student makes multiple evaluations of each teacher, you can use the average score (or total score) by each student of each teacher.

Charles

hello. I am writing from Mongolia. First of all, thanks for your help. I have been doing the survey that has 5 foint likert scale. the survey is Student satisfaction for teachers. my university has 3000 students and all students give answer to 130 teachers. How to calculate cronbach’s alpha?

Good to communicate with someone from Mongolia.

Enter the Likert scores in a range with 130 columns and 3000 rows and then either follow the approach from Example 4 or use Real Statistics’

CRONALPHAfunction or the Cronbach’s alpha option from theReliabilitydata analysis tool.Charles

Thank you very much for answering my question.

Dear sir,

i had done a survey on questionnaire which contain 3 parts (A,B,C). I’m using cronbach alpha for my questionnaire. I had calculated those 3 parts and only part A and C are reliable, whereby the value part A ( 0.798), Part B ( 0.642) ,Part C (0.857). So what should I do with those result and are they reliable?

Adele,

The reliability ratings for Part A and C are quite good. Even Part B is not bad. The typical approach now is to see how Cronbach’s alpha changes by removing one question (as described on the referenced webpage). If say removing question Q4 from Part B increases the reliability for Part B to .7, then you should consider eliminating this question from Part B or changing it.

Charles

TQ sir for answering my question. I would like to remove the question but most of the score of all the questions in Part B are around 0.6. So should i remove all the questions?

Adele,

You probably don’t want to remove all the questions. You might want to think about changing the wording of all the questions a little. Alternatively, you can do nothing since .6 is a little on the low side, but acceptable to some.

Charles

Pls sir how did you calculate variances in row 17 and 21. will you mind using Figures from the table?

I really want to know this cronbach alpha

This is already explained in Figure 2 and the text between Figures 1 and 2. What else are you looking for?

Charles

does anyone know what the lower bound for the number of observations using the alpha is? i’ve heard it’s about ten times the number of items, however can’t be sure at all… many thanks!

Vince,

The Real Statistics Resource Pack provides a data analysis tool which calculates the sample size required for Cronbach’s alpha. See the following webpage for details:

http://www.real-statistics.com/reliability/cronbachs-alpha/cronbachs-alpha-continued/

Charles

I am research scholar. I need to know when to calculate Alpha. E.g For testing reliability, can I go with 10 sample units and the value of Alpha is 0.897. Would it be fine???

Give me reply somebody.

I don’t completely understand your question, but if you are using Cronbach’s alpha to test the reliability of a questionnaire based on 10 subjects, an alpha of .897 would generally indicate a high level of reliability.

Charles

Dear sir,

I want to use a test with high zero frequency in scoring , if the answer is rarely or sometimes or never then the score is equal zero (represent the same number) ?! so Cronbach’s alpha is low ?! but if I insert the data as, never represent=6 , rarely=5, sometimes=4,usually=3, often=2, always=1 then Cronbach’s alpha will increase !!

can I do that only to calculate Cronbach’s alpha ?! I mean changing the scoring only for the pilot study?!

Best,,

I just recalculated Example 4 on the referenced webpage using a Likert coding of 0 to 6 (instead of 1 to 7) by reducing each code by 1. The calculated value of Cronbach’s alpha did not change. Is this the sort of thing you are referring to?

Charles

Hi Charles,

I am so glad to find your website. Your demonstration of how to calculate Cronbach alpha is very relevant to my research. May I ask you one question?

I used a spelling test from a published study for my own research and I want to report the internal consistency reliability of this test based on the data I have collected. The spelling test examines how students spell four English vowel sounds. In such case, do I need to calculate Cronbach alpha for each target vowel sound? Is it appropriate to calculate Cronbach alpha for all items (all sounds) and report internal consistency reliability of the whole test?

Thank you so much for helping me.

If the test is for spelling of vowel sounds then one Cronbach’s alpha should be sufficient, but if you suspect there may be differences between vowels then I would compute separate Cronbach’s alphas for each vowel sound. I would do both and explain the results.

Charles

Thank you very much for answering my question.

Hi Charles,

Sorry I have one more question. I have four groups of participants (who have different characteristics) in my study and the same spelling test was administered to all groups. Do I need to calculate Cronbach’s alpha of the test for each group?

Thank you for helping me again.

I would do both. If the difference in the characteristics of the participants changes the reliability of the test, then this should become evident in the Cronbach’s alpha values.

Charles

Thank you!

Good day sir.

In my experiment survey of Y/N(1-0) questions, where all 3 questions got a perfect score of 1 from 15 respondents, I was not able to compute the alpha. The value of all the variances is 0.

The final result is k=3, eVar=0, var=0.

What could it mean?

Is it also possible to get a perfect score of 1 or 0 for the alpha? What would each result mean?

Thank you.

-K

K,

Unfortunately, in these circumstances Cronbach’s alpha is undefined because of division by zero. The good news is that all the respondents answered the same way, it is pretty easy to make evaluate the effectiveness of the survey even without using Cronbach’s alpha.

If on each of the three questions the subjects didn’t have identical scores, but if each subject got the same score on all three questions, then Cronbach’s alpha would be 1.

If you assign all scores of 0 or 1 at random, then Cronbach’s alpha is likely to be low (even negative).

Charles

Sir,

I want to know which input or inputs (out of 22) has more impact on output 1

which input/ inputs (out of 22) has more impact on output 2

which input/ inputs (out of 22) has more impact on output 3

which input/inputs (out of 22) has more impact on output 4

Raman,

Sorry, but the term “more impact” is unclear to me.

Charles

Hello Sir,

Firstly i want to say thanks for helping me to compute Cronbach’s alpha Value.

I go through almost all study material provided by you. But still in confusion which more statistical tools can be implement on my questionnaire data.

There are 22 input and 4 output.

I collected data for every input with respect to every output on likert scale

Q1 Rate the effect of Input 1 on Output 1

(Importance scale for every question)

(Extremely (Very Neither Imp (Very (Extremely

Important) Important) Nor UnImp. Unimportant) Unimportant)

1 2 3 4 5

Q2 Rate the effect of Input 1 on Output 2

Q3 Rate the effect of Input 1 on Output 3

Q4 Rate the effect of Input 1 on Output 4

Q5 Rate the effect of Input 2 on Output 1

Q6 Rate the effect of Input 2 on Output 2

Q7 Rate the effect of Input 2 on Output 3

Q8 Rate the effect of Input 2 on Output 4

Q9 Rate the effect of Input 3 on Output 1

………… and so on

Q85 Rate the effect of Input 22 on Output 1

Q86 Rate the effect of Input 22 on Output 2

Q 87 Rate the effect of Input 22 on Output 3

Q88 Rate the effect of Input 22 on Output 4

The no of respondent : 100

Thanking you in anticipation

Raman,

If I understand correctly, you have 88 questions using a Likert scale of 1 to 5 and 100 subjects. This is just like Example 4 of the referenced webpage but with 100 rows and 88 columns. You can use the CRONALPHA function or the Reliability data analysis tool to calculate Cronbach’s alpha.

Charles

Sir,

I have computed the value of Cronbach alpha.

But my query is now let me know which more tools such as Z test, correlaion, regression, factors could be used. The objective of my survey is to examine the impact of input factor in output.

Raman,

The phrase “examine the impact of input factor in output” is too vague. You need to create a hypothesis and then decide which test best tests this hypothesis.

Charles

Sir,

Thanks for replying

i want to know how many input parameters significantly influence output1; how many parameters significantly influence output2,parameters significantly influence output3, parameters significantly influence output4.

Raman,

What do you mean by “significantly influence”?

Charles

Hello,

Thank you for this very helpful site! I am curious about the use of Excel’s VarS function (instead of VarP). I notice you use VarP above, and I’m wondering why that is. If I understand Excel’s explanation, VarS would be used for a sample whereas VarP is used for a population. Wouldn’t most, if not all, of the scales we may develop require the formula for sample variance? Can you help me understand your choice to use VarP so that I can distinguish it’s use in my application of Cronbach’s Alpha for scales I am creating?

Thank you in advance,

Brenda

Brenda,

This is more of a mathematical issue. If you replace Var.P by Var.S throughout you will get the very same answer (since the n or n-1 terms in the denominator cancel out).

Charles

Oh yes, of course! Thank you!

Thanks Charles for ur kind reply. I want to check the influence of input parameres on the output parameters of a system. for this questionnaire study was performed.

the data collected table has 100 rows and 15 column.

I got the single value of Cronbach’s alpha. If there need to compute the value if Cronbach’s alpha for every column. If yes then how.. Kindly do needfull

Raman,

In the examples I have given on the referenced webpage, the columns correspond to “items” in a questionnaire (these are variables). Not only don’t you need to compute Cronbach’s alpha for each column, but such a calculation would yield an error since Cronbach’s alpha requires more than one item. On the website, however, Cronbach’s alpha on one column corresponds to Cronbach’s alpha on all the item leaving out that one column. This tells you the influence that specific item has on the entire questionnaire.

Charles

Thanks a lot

For example 1, can you explain why cells B20 and B21 do not have the same value? I’m trying to follow the formulas, but it seems like M17 in B20 would have the same value as the result of the formula for B21. Thank you!

Sammie,

B21 contains the variance of the row sums, i.e. VARP(M4:M15), while B20 contains the sum of the variances of the column sums, i.e. SUM(B17:L17). There is no reason why these should be equal.

Charles

Hi Charles,

My querry regarding Cronbach’s alpha values. No of respondand =100, The questionaire having 5 point scale ans consisit of 22 input and 4 output facors. I used the procedure expalined by you in figure 1. I got value. But some researcher are showing Cronbach’s alpha values for every input and output factors. If there need to compute the value if Cronbach’s alpha for for every input and output factors. If yes then how.. Kindly do needfull

Raman,

I don’t know what you mean by input and output factors in the context of Cronbach’s alpha.

Charles

Hi Charles,

I want to ask you about cronbach alpha.I have questionnaire which in this questionnaire have Likert type scale (1-4) and question yes/no.In questionnaire also have question like gender,marital status etc.How to get the cronbach alpha value using these data?Thank you

Hi Azili,

The idea of Cronbach’s alpha is to measure the internal consistency of the questionnaire. If the yes/no questions are measuring something different from the Likert questions (e.g. the yes/no questions are testing the subject’s knowledge and the Likert scale questions are determining their satisfaction) then there may be no point in calculating one Cronbach’s alpha; better to calculate two or more values.

If instead the yes/no questions are measuring something similar to the Likert questions then perhaps you can decide to treat the yes/no questions in a manner similar to the Likert scale questions. E.g. if the Likert scale questions measure satisfaction (with 0 very dissatisfied and 4 very satisfied) and yes = satisfied and no = not satisfied, then perhaps a coding like yes = 3 and no = 1 will work. In most cases, I suspect that the yes/no questions are measuring something different from the Likert scale questions and so you need to calculate multiple Cronbach alphas.

Charles

Hi Charles,

Hope you can help. We did a survey on identifying comparison in students’ attitude towards a list of 10 sub-factors (5 point Likert scale) on advantages of using a tool. The Cronbach alpha for each of the 2 colleges (A and B) with 2 columns (A and B) is over .9 which should reflect good reliability. My question is, if we then look into overall gender comparison analysis on students of the same two colleges using the same sub-factors with 2 columns (male and female), is it necessary to evaluate the Cronbach’s alpha for the gender comparison or can we assume that the reliability is the same?

Thanks in advance.

Kent,

If you suspect there is a gender bias to one of the questions, then I would evaluate Cronbach’s alpha for each gender. If one question seems to lower the score for one gender, then I would either reword or eliminate that question. If your goal is to compare the results by gender then you want to use a different test (e.g. t test, Mann-Whitney, etc.).

One other comment. The value you obtained for Cronbach’s alpha was high, but since it is over .9 it may be too high, meaning that the questions may be too similar and are just testing the same things.

Charles

Tqvm Charles for your response.

Yes, we do conduct the t-test for comparison. I wonder if it is necessary to provide Cronbach’s alpha value to indicate good reliability and internal consistency when we are using the same set of questions for college and gender comparisons. Is it a necessity to provide Cronbach’s alpha value to validate good reliability before doing a comparison between colleges or gender?

Thanks in advance.

Kent,

If you have gotten high values of Cronbach’s alpha in your test sample it is not necessary to repeat Cronbach’s alpha for new samples, unless you suspect there is something different about the new sample. As I stated in my previous response, if you suspect there is a gender bias, not in the responses per se, but in the internal consistency (perhaps because one question is misinterpreted by men or women), then you should repeat Cronbach’s alpha. In any case, since you will be testing the data anyway you probably should run CRonbach’s alpha just to make sure there is nothing strange going on.

Charles

Tqvm, Charles.

You have been a great help.

Take care.

I am trying to determine the reliability of an exam in which there are 25 multiple choice exams with a variable number of response choices (4, 5 or 10). Blank answers are given 1/4 credit. What is the most appropriate way to measure the exam reliability?

exams should have read questions

Mark,

The number of response choices shouldn’t really matter since each question is scored 1 for correct or 0 for incorrect. However things are more complicated with 1/4 credit for a blank answer. Perhaps one way to address this is to assign a score of 0 for incorrect, 4 for correct and 1 for blank and then use Cronbach’s alpha.

Charles

love the info. thanks so much.

so nice to have experts share their knowledge on the net.

what about sample size shoud be taken to test reliability having 20 questions with 7 piont scale likert type

Nasir,

To answer this question you need to spply additional information such as the expected/minimum value of Cronbach’s alpha, minimum power, etc.

Charles

Thank you, thank you, thank you for this clear and logical explanation of how to apply Cronbach’s Alpha to a Likert scale.

Hi I have another query to add to this. I have run the Cronbach’s alpha for 2 of the individual manager’s sets of results for one of the scales and a set of data I would have thought should have shown far lower inter correlation than the other came out the other way round. Is it because I only have about 6 individuals results – which is too low?

Yvonne,

It is difficult to answer your question without seeing the data, but as I said in my previous response it doesn’t sound like Cronbach’s alpha is the correct measurement to use in any case.

Charles

Hi firstly thank you so much for this website it has been so helpful. I am doing an MSc just now and am looking at manager’s self-awareness and how closely their followers/direct reports assessment of their leadership styles agree. I have only 7 managers and 26 direct reports results.

I have a few queries from when I apply Cronbach’s alpha – should I be looking at the manager and their direct reports results as individual cronbach’s alpha? Or can I combine all the data as one set of 33 results? Also when I first ran Cronbach alpha on the 6 questions which made up one scale, I got a figure of >1. I worked out this was because the missing data points must be being used as 0s. I first excluded the whole column of data (ie 3 of the criteria for the scale) and got a figure, then I ran it again removing the people with the missing data – and got a similar figure to that of removing the question. Which is correct way of doing this? or how else should I deal with missing data? Also, looking at an earlier response Cronbach’s alpha may not be very accurate/reliable with such a small sample? Any other ideas of how I could analyse the data? The purpose of my MSc question was to discover if the manager’s view of their leadership style matched that of their direct reports. So possibly I do need to do cronbach’s alpha on each manager – rather than using it as one data set?

Yvonne,

The current implementation of Cronbach’s alpha doesn’t accept any missing data. If you remove an item, this item won’t be included in your analysis. If there is a lot of missing data for one item you might consider removing this item, but in this case you should try to understand why there is so much missng data for this item. Otherwise, I would tend to remove the data for subjects with missing data. Of course this approach only works if there is a small amount of missing data. Otherwise you should consider imputing values for the missing data.

Based on your description, it seems that you are interested in measuring the agreement between raters. Usually intraclass correlation, Cohen’s kappa, Fleiss’ kappa, etc. are used for this (and not Cronbach’s alpha). You can get more information about these measurements on the webpage http://www.real-statistics.com/reliability/

Your situation is the agreement between one rater and a group of raters. This discussed in Chapter 4 of http://orbi.ulg.ac.be/bitstream/2268/39575/1/vanbelle-thesis-5-5-2009.pdf

Charles

Charles,

I have done the split-half methodology against our set of multiple-choice questions; and thus far, the results look good. Since students may not get the same question (although all questions measure the same concept(s)) on their specific exam, I had to sample each conceptual question into groups of 30.

For example, I would have thousand rows with the columns of NumQuestionOffered=30, and NumQuestionCorrect=X

Provided I have the Odd/Even Anlysis Means, StdDevs, etc. ..is it possible to use that data to produce a standardized Cronbach’s Alpha?

Sorry Martin, but I don’t completely understand what you are trying to do. I see that you are trying to combine questions, but I don’t understand how the split half accomplishes this.

Charles

I have 6 items with SA, A, SD, D, how do I apply cronbach alpha with these figures.

Udo,

Just assign the values 4 to SA, 3 to A, 2 to D and 1 to SD and use the approach described in the referenced webpage.

Charles

Thanks for the great resource pack and website. My question: While using Fleiss’ Kappa to measure rater agreement, I experimented with various agreement scenarios but now have trouble interpreting the kappa. With 11 subjects and 5 raters on a 5 point scale, I set everyone’s agreement on the middle of the scale (e.g., 3 on a 1 to 5 scale). Kappa would not calculate until I moved 1 score into another column. So, with all raters=3 for all subjects except one, kappa calculated to -0.01852. Intuitively, I expected a number close to 1. I’m left wondering if the answer is 1 minus the calculation.

I appreciate your help.

Frank,

Interpretation of Fleiss’ Kappa seems to be very difficult and depends on many factors (sample size, variability, bias, etc.). See the following article for a description of some of the problems. http://digital-activism.org/2013/05/picking-the-best-intercoder-reliability-statistic-for-your-digital-activism-content-analysis/

As mentioned in this article the problem you have identified is inherent in the measure (and not my implmentation).

I am about to look into Krippendorff’s Alpha-Reliability as an alternative measure. The following is an article that explains this measure:

http://repository.upenn.edu/cgi/viewcontent.cgi?article=1043&context=asc_papers

Charles

Many thanks for your reply. I look forward to new your new insights. – Frank

What Statistical tool can best be used to test hypothesis of significant INFLUENCE of one variable over the other. the two variables have separate instruments measuring them. Is it T-test, Correlation, Regression, Chi-square, or which

Thanks

Influencers is a term used in association with residuals in regression. See the webpage http://www.real-statistics.com/multiple-regression/outliers-and-influencers/ for more details.

Charles

good afternoon sir. I am conducting a comparative study to assess the developmental milestones of children in rural and urban areas. I am using checklist as a tool. My tool have three categories 6 years,7 years,8 years. Each categories has different questions. 6 years contain 6 question. 7 years category contain 7 questions. 8 years category contains 7 question. I want to check the reliability of tool. Than how I will check the reliability of tool

Good afternoon sir.

my tool is checklist. It has three catageries 6 Year(1 catg),7 Year(2catg),8 year(3 catg). 1 catg contains 6 items,2 catg and 3 catg contains 7 items. How i will calculate reliability.

thank you

Sona,

Unfortunately, you haven’t provided enough information for me to make any recommendations.

Charles

This a great websites I have gone through several books and websites nowhere I got all these information My sincere respect to you.

Regards

Shiba

Asst Prof

Shiba,

I am very pleased that you find the website valuable. I am trying to create an easy-to-use, yet powerful resource that is available to everyone who needs to learn, teach or use statistics.

Charles

There are typos in your equations for x_0 and t_0. You should be summing over x_j and t_j (i.e.: $x_0 = /sum_{j=1}^k x_j$ and $t_0 = /sum_{j=1}^k t_j$), respectively.

Lou,

Thanks for identifying these typos. These sorts of mistakes can really cause confusion for people who are less familiar with the notation, and so I really appreciate your comment. I have now made both corrections.

Charles.

Charles:

I am conducting multi-variable regression using 5 independent variables to predict the impact on transportation costs as a proportion of income in 75 metropolitan statistical areas. The data includes employment rates, population density, and value of exports. Is Chronbach’s Alpha an appropriate test for reliability or should I be using a different type of test? If is is appropriate, should I be including the dependent variable in the calculation? Thank you very much.

Reggie,

I can’t see why you would want to use Cronbach’s alpha in this case. What is the objective of the reliability test you want to pperform?

Charles

Charles, thank you so much for this article; it is very helpful for a math novice like myself. I would be grateful if you could assist me with some questions regarding the calculation of the Cronbach Alpha.

We have a test which deals in direct assessment. To be brief, I would like to calculate the reliability of our testbank broken down into topical groups. While our testbank will contain about 300-400 questions per topic alone, each student is only presented with 10 random questions. I have a population of 800. Does it matter if the questions (all multiple choice) are different for each student? Can I still create 10 columns with 800 rows?

My assumption tells me I cannot, because Q1 for student 1 may have gotten it correct, and student 2 and 3 may have also gotten it correct, however, Q1 for student 2 and 3 could be the same, or, most likely entirely different (although all the concepts are the same).

Please advise

I don’t know how to calculate Cronbach’s alpha in this situation.

Charles

hie,

what percentage of respondents do i need when using cronbach alpha during pretest? where can i get this reference. a friend told me that i need 10% of the respondence, is it true please help.

Maggy,

I have seen a number of websites that have information about the minimum sample size for Cronbach’s alpha. I used the following search on google: “minimum sample size for cronbach alpha”

I have also seen the following paper which you may find helpful:

Research in Nursing & Health, 2008, 31, 180–191

Considerations in Determining

Sample Size for Pilot Studies

Melody A. Hertzog*

Charles

Hi Charles

Thank you very much for the resource you provide.

My question is the following: I usually run alpha on a very limited sub-sample (about 10 cases) before delivering any questionnaire in order to asses the consistency of it; btw I have no reason for choosing this number, 10 cases. Is there any rule suggesting a minimum number of cases or a minimum ratio cases/variables to have a ‘solid’ alpha?

Giacomo

Giacomo,

I have seen a number of websites that have information about the minimum sample size for Cronbach’s alpha. I used the following search on google: “minimum sample size for cronbach alpha”

I have also seen the following paper which you may find helpful:

Research in Nursing & Health, 2008, 31, 180–191

Considerations in Determining

Sample Size for Pilot Studies

Melody A. Hertzog*

Charles

Thank you again!

Cheers

Giacomo

Charles,

First, let me thank you for these excellent tools. They provide wonderful access to statistical results related to educational outcomes.

I run Excel 2010 and have the downloads installed. However, I can run cronalpha, but not with the k switch. Nor can I run calpha. What do you think is missing, misplaced, or uninstalled.

Roy

Roy,

I introduced these capabilities quite recently. If you are using a version of the software prior to Release 3.0 these capabilities are not included. I suggest that you download the latest version of the software.

Charles

i need draw backs of cronbach alpha test.

why cronbach is important for internal consistency?????

Cronbach’s alpha measures internal consistency. Internal consistency is important because it determines whether a questionnaire or some other measuremnt tool is measurin the same thing. If for example you design a questionnaire to determine whether someone is bipolar, you want to make sure that all the questions are measuring the same thing,

Charles

A couple of drawbacks:

– Cronbach’s alpha measures internal consistency, but the more items the higher Conbach’s alpha will be even though the internal consistency isn’t any higher.

– Redundant items (e.g. the same question phrased slightly differently) will increase Cronbach’s alpha

Charles

Dear Charles,

Really hope you can solve my problems.

My questionnaire consist 3 parts which are

1. Knowledge in action research – Using test form (5 multiple choice answers and 5 open-ended questions)

2. Skills in implementing action research – Using 5 point likert scale

3. Attitude towards action research – Using 5 point likert scale

My problem is how to run reliability test for part 1. Because in this part have few difference types of questions.

Example:

1. multiple choice (A , B , C , D )

2. list out 4 data collection techniques

3. pls arrange 10 steps of implementing action research using number 1 to 10.

My friend advise me to run Kuder Richardson, which one more suitable?

If use Kuder Richardson to get reliability for part 1, then how about the overall reliability for the questionnaire? How to run the test?

Anyway thanks a lot.

The multiple choice questions can be coded as 0 for a wrong answer and 1 for a correct answer. You could code the 4 data collection techniques as four questions with 0 for an incorrect technique and 1 for a correct tecnique. I am not sure how you want to code the 10 steps question as right or wrong (this is necessary if you want to use Kuder Richardson).

You can use Cronbach’s Alpha. If all the scores are 0 or 1 the result for Kuder Richardson is the same as for Cronbach’s alpha (the referenced webpage). You can also look at the webpage http://www.real-statistics.com/reliability/kuder-richardson-formula-20/.

What is most important is that Kuder Richardson (or Cronbach’s alpha) is calculated only from questions that are measuring/testing the same thing.

Charles

Thank you very much for your valued information given above. I really learn a lot. Only I have one question. If I have three different companies answering same questions? Can I measure the internal consistency by conducting Cronbach’s alpha? Even ratters from company A have different circumstances than B or C?

Faisal,

Cronbach’s alpha doesn’t work with only one question, although there are some other procedures for dealing with one question. I gave a link to one of these in a previous comment. I wouldn’t expect much with such a limited amount of data.

Charles

Dear Charles,

Thank you very much for your prompt replay. I mean I would like to ask only one question. I am sorry it was my fault not describes my question very well. I have 10 groups with 43 factors and I have 36 response from three different companies rate my factors. If I have three different companies answering same questions? Can I measure the internal consistency by conducting Cronbach’s alpha? Even ratters from company A have different circumstances than B or C?

Dear Faisal,

Yes, you can use Cronbach’s alpha to create one value of alpha (or three values, one for each company). It would be interesting to see whether you get different values for the three different companies.

Charles

I have a mixed set of questions in my questionnaire where some questions are based on 5 point likert scale and some are single response questions not dichotomous. So, how should I calculate the cronbachs-alpha value for all set of questions in my questionnaire.

Sadhna,

The important thing is to calculate a separate Cronbach alpha for each group of questions that are testing the same thing. Stated the other way around, you shouldn’t calculate Cronbach’s alpha for a set of questions that are testing different things.

Regarding your specific request, can you give me some idea of the single response questions that you are using?

Charles

Got your point, thanks. lets say I have question 1. Which parameter do you look for while selecting the service provider?a)call charges b)network availability c)Customer care services d) offer and schemes e) any other. For these types of questions, how to calculate cronbachs alpa value.

please, how do I compute the correlation

between a particular item and the

sum of the rest of the items.

Say the the data is in range A1:F50 and you want the correlation between the data in column E and the sums of the rest of the items. Now place the formula =sum(A1:F1)-E1 in cell G1, highlight range G1:G50 and press Ctrl-D. The result that you are looking for is calculated by the formula =CORREL(E1:E50,G1:G50).

Charles

Dear Charles,

Is it possible to determine the reliability of a survey questionnaire even if there are no established correct answer since it deals with their opinions? Should it be possible, what are the steps in computing it using the spreadsheet?

Riza

Dear Riza,

Yes. E.g. if the answers are in the form of a Likert scale then you could use the Likert scale score to calculate Cronbach’s alpha.

Charles.

.

Hi there,

I have a questionnaire which contains:

1) Yes/No items

2) Likert-scale items

Can I use Cronbach? If yes, how do I perform the coding for 1) & 2) in SPSS?

Appreciate your advice.

Debbie

Hi Debbie,

The important thing is to calculate a separate Cronbach alpha for each group of questions that are testing the same thing. Stated the other way around, you shouldn’t calculate Cronbach’s alpha for a set of questions that are testing different things.

I would not mix Likert scale questions (rate your opinion of Columbus from 1 to 5) with knowledge questions (did Columbus discover America?).

I don’t use SPSS and so am not familiar with the coding in SPSS. For the Real Statistics Resource Pack you would code a Likert scale of 1 to 5 with the numbers 1 through 5. You would code True/False (as well as multiple choice questions) with 1 for the correct answer and 0 for the incorrect answer.

Charles

Dear Charles,

Thank you so much! Much appreciated.

Debbie

Hello Charles,

I wish to use cronbach to demonstrate some internal consistency of a survey. If this has different sections and the sections each have mixtures of types of questions e.g, likert,dichotomous,multiple choice..how do I test for cronbach please? Thanks

The important thing is to calculate a separate Cronbach alpha for each group of questions that are testing the same thing. Stated the other way around, you shouldn’t calculate Cronbach’s alpha for a set of questions that are testing different things.

In my view there is no problem mixing true/false and multiple choice questions provided the above paragraph is adhered to.

For multiple choice questions of the form “Select one of the following four choices a, b, c or d” if say (c) is the correct choice then use the value 1 if the person selects choice (c) and use the value 0 if the person selects a, b or d.

Questions of the form “Select one or more of the following four choice a, b, c or d”, can be viewed as four True/False questions, namely choice (a) is correct (True or False)?, choice (b) is correct (True or False)?, choice (c) is correct (True or False)? and Choice (d) is correct (True or False)? If say the correct choices are (a) and (d), and a person chooses (a) and (b) instead, then score 1 for question (a), 0 for question (b), 1 for question (c) and 0 for question (d).

I would not mix Likert scale questions (rate your opinion of Columbus from 1 to 5) with knowledge questions (in which year did Columbus discover America?).

Charles

Hi, I am currently working on my research. I face problem when i generating Cronbach Alpha. My questionnaire contains such question :

If yes,please answer Section B.

If no, please answer Section C.

It’s quite complicated, how to obtain the Cronbach Alpha value after i key in all?

If answer section B, what i need to fill for section C items? With missing value or without missing value?

Need a guide to complete it as I running out of time.

Kyan,

Perhaps the easiest approach is to calculate Cronbach’s alpha for the questions in section B (ignoring the people who didn’t answer the questions in section B) and calculate a separate Cronbach’s alpha for the questions in section C (ignoring the people who didn’t answer the questions in section C).

Charles

Dear Charles,

I have followed the chat, however, still did not get how to calculate alpha with multiple choice questions.

What do you mean by “correct (1)/incorrect(0)”? For instance, my questionnaire contains some questions, beside of Likert scale questions, with 5 sentences and the respondent is asked to mark those sentences which he associate with the particular subject. That means many different combinations could be marked.

Also there are such questions which require to range 5 sentences from most favourable to least favourable – also the answers could vary alot. How to cope with such questions while calculating alpha?

Thank you very much

Dear Darius,

For multiple choice questions of the form “Select one of the following four choices a, b, c or d” the approach that I suggested should work, namely if say (c) is the correct choice then use the value 1 if the person selects choice (c) and use the value 0 if the person selects a, b or d.

Questions of the form “Select one or more of the following four choice a, b, c or d”, can be viewed as four True/False questions, namely choice (a) is correct (True or False)?, choice (b) is correct (True or False)?, choice (c) is correct (True or False)? and Choice (d) is correct (True or False)? If say the correct choices are (a) and (d), and a person chooses (a) and (b) instead, then score 1 for question (a), 0 for question (b), 1 for question (c) and 0 for question (d).

I am not sure how best to handle questions of the form “List the following five choices from most favourable to least favourable”. They are similar to having five questions where you rate each on a Likert scale of 1 to 5, but the difference is that you can’t use the same Likert score twice. Perhaps someone else has an idea about how to handle these types of questions

Charles

I´ve just discovered Real Statistics, and so far it looks fantastic!

I´m looking to automate lots of calculations on a complex data set.

Is it possible to make the formula only calculate certain questions (for a subscale)?

How does the formula handle missing items/participants?

Kind regards,

Martin

Martin,

Real Statistics has a lot of useful statistical functions and data analysis tools. I am also adding new features all the time.

There are various functions for selecting only certain data from a data range. These consist of standard Excel capabilities (

Data > Filter, etc.) and Real Statistics capabilities (e.g.Extracting Columns from a Data Rangedata analysis tool andExtractColfunction).Many of the statistical functions ignore any rows with missing data, but the

CRONALPHAdoesn’t have this capability. You need to first remove any missing data (e,g, by using the Real Statistics functionDELROWBLANKorDELROWNonNumor via theReformatting a Data Range by Rowsdata analysis tool) and then apply theCRONALPHAfunction to the reduced data range.Charles

Just wanted to say what a fantastic resource you have put together here. I’m an educational researcher from a purely Arts background.

I put together a Likert survey, made up of 6 root statements (or factors?), and chose to analyse 3 of them using this method. I had 79 respondents and used a 5 point scale. The first factor consisted of 7 items and the second and third of 4 and 3 respectively. The spreadsheet returned scores .88, .84, and .79. I am well pleased!

I may even have a go at using your site for Factor Analysis and go a step further…

Much, much appreciated.

Chris, New Zealand.

Chris,

I am very pleased that you found the site useful. Interestingly enough the original impetus for creating the site was that I wanted to perform Factor Analysis in Excel. Once I figured out how to do this I decided to share what I had learned with others who might want to do real statistical analysis using Excel.

Charles

Hi Charles, firstly thank you very much for providing this informative platform for everyone. Your effort is greatly appreciated.

Anyway, i do have some issues regarding alpha cronbach. Hoping that you can give me some clarification. Issues are stated as follows;

(i) In one full set of questionnaire, i have 42 items/ questions measuring different and diverse concepts/ factors. In another word, i cannot put all the items together at one go for Cronbach test (as this will lower the cronbach value as they are measuring different things), so i separate them into few groups say 8 groups of factors (consists of 3-5 or 2-3 items). So theoretically, each group i have different value of Cronbach if i executed cronbach for all the items. The question is do i need to execute the cronbach test for ALL items or groups? Is it necessary to ‘cronbach’ all the items? Can i just ‘cronbach’ just one group?

(ii) In exension of above issue, as you mentioned earlier, in order to use Cronbach, you have at least 2 items right? but we know that the lesser the items, the lower alpha cronbach will be. Say i have low cronbach value for 2 items, but these two items are valid to be in questionnaire just the reported reliability is low, so, do i need to do the cronbach test for that group? or i just ignore it? or there is a better alternative?

(iii) For your information, another issue is there are some items, like 2-3 items they are unique (maybe i have to say, they are not ordinal data, they are nominal data- yes or no etc, meaning to say, they are representing individual factor, so cronbach cannot be used right? is there alternative to replace the cronbach if this is the case? One more issue is multiple choices questions, i have about 8 of them, so, how can i establish reliability of them, is that possible?

Thank you very much

Gabriel,

(i) You only need to calculate Cronbach’s alpha for the group(s) you are interested in.

(ii) As you know, all other things being equal, in general the more questions you have the higher the value of Cronbach’s alpha will be. For this reason the value of Cronbach’s alpha for two items (i.e. two questions) might seem low compared to a group with more items. There may be better alternatives to Cronbach’s alpha for two items, but I am not aware of them.

(iii) Cronbach’s alpha handles true/false and multiple choice questions with no problem. You code these with 0 for the wrong answer and 1 for the correct answer (as in Example 1 of the referenced webpage).

Charles

Hi,

for my thesis I conducted CRONBACH alpha on several measures- the core questionnaire is the Five Facet Mindfulness Questionnaire (FFMQ). My sample is 265 Norwegian University students.

These are the scores I need to explain: FFMQ (0,582) where the individual facets never reach even 0.5, for the Perceived Stress questionnaire (0,345) and for AUDIT (0,396).

My supervisor thinks these are too low and would like me to explain why, and thinks we cannot be sure what the instruments were actually measuring. He wants me to give reasons for this.

Also, the CORRELATION between measures was low even though in the past it usually shows to be high. I am not sure what was different now, and what both a low cronbachs alpha and correlation could indicate.

I have found some reasons that could explain it like Participant Fatique.

I would truly appreciate any help as my thesis defence is due very soon and I am still stuck at this problem.

Best,

MJ

MJ,

Since this is a standard test, I assume that your supervisor finds that your results are too different from the typical results for the test. You mention that a possible reason is Participant Fatigue, but you need to ask yourself why would students taking this questionnaire have higher levels of fatigue than the typical respondents to this questionnaire? In general, I can only think of two categories of reasons for the discrepancy: (1) the people taking the test are different from the typical people taking the test (different culture, problems with language, etc.) or the environment where the questionnaire is administered is different (too noisy, poor lighting, etc.) or (2) you aren’t measuring the the same thing as is typically measured — e.g. if I remember correctly the FFMQ has 5 different dimensions (factors), in which case you need to calculate Cronbach’s alpha not on the whole questionnaire but for the questions in each of the five dimensions separately.

Charles

I have multiple choice question with one correct answer

How can i get internal consistancy for my questionnaire, should i use similar method in spss to get chronbach alpha vlue for likert scale for my 15 question of MCQ??

First in coding a b c d with 1.2.3.4

But now..i coding them 1 for correct answer and 0 for uncorrect answer

So what should i do next??

Farah,

When using a Likert scale (such as 1, 2, 3, 4) you should use the appropriate value 1, 2, 3 or 4 for each question, but multiple choice questions are not like questions using a Likert scale. When you are analyzing multiple choice questions you should code 0 (incorrect) and 1 (correct). After the coding just following any of the approaches described on the referenced webpage to obtain Cronbach’s alpha.

Charles

A lot of thank for replied my massage..:)

By the way, what do you mean by following any approaches described on reference webpage to obtain cronbach alpha?

I hv difficulties to analyze my data to get alpha value..

What should i use..anyway..any suggestions?

Farah,

On the referenced page I showed three ways of computing Cronbach’s alpha:

1. Directly using the definition of Cronbach’s alpha (as in Example)

2. Using ANOVA (as in Example 3)

3. Using the CRONALPHA function found in the Real Statistics Resource Pack (as in Example 4)

Charles

Thank you very much charles:)

I still dont understand about why should we change into 6 instead of 2…

N Q20….”Q20 in our example. Thus if a response to Q20 is say 2, it needs to be scored as 6 instead of 2 (i.e. 8 minus the recorded score).”

And why must be 8?

Im very confius…

Hope u can help me…

Thanks

Farah,

Suppose you pose the following four questions and ask your respondents to give a rating of 1 to 7 for each, where 1 is the weakest response and 7 is the strongest response:

“Do you like pie?”, “Do you like steak?”, “Do you like green beans?” and “Do you dislike pizza?”

The fourth question is reverse phrased: “dislike” instead of “like”. A response of 7 to the first question means that the respondent really likes pie, while a response of 7 to the fourth question means that the respondent really dislikes pizza. This last response is equivalent to a response of 1 to the question “Do you like pizza?”. So in order to “compare apples with apples” (pun intended) you pretend that the last question was “Do you like pizza?” and use a score of 1 instead of 7.

In a similar way, a response of 6 to the question “Do you dislike pizza?” is equivalent to a score of 2 to the question “Do you like pizza?” Note that with a Likert scale of 1 to 7, the reverse coded score is always 8 minus the actual score (here 8 is 7+1). Similarly, if the Likert scale is 1 to 5, the reverse coded score is 6 minus the actual score (here 6 is 5+1).

I hope this helps.

Charles

Dear Charles,

I have made a series of multiple choice questions with 70 items and gave them to a group of ten for piloting but it is really hard to estimate Cronbach’s alpha. I know that it should be more than 0.7 but I don’t know how. Please help me. Thank you so much in advance.

Sincerely,

Marzieh

Dear Marzieh,

If the person answered the question correctly they score 1 for that question. If they answered it incorrectly they score 0 for that question. Then use the approach in Example 1 of the referenced page or the CRONALPHA formula to calculate Cronbach’s alpha.

Charles

Hi,

I have 4 questions of type Likert Scale ( each of them contain 4 questions describing the main question).

As far I saw that cronbach alpha is for internal consistency, or it shows how good items are related to describe the main question.

I want to measure inter-rater agreement. Is Cronbach Alpha the right metric for doing this?

Thank you

Hi Ermira,

Depending on your specific requirements, you could use Cohen’s kappa, Weighted kappa, Fleiss’s kappa, ICC or Kendall’s W. These all measure inter-rater agreement and are described in the website. See http://www.real-statistics.com/reliability/ for more details.

Charles

Hello,

Thank you for the explanation and the answers that you gave, it is very helpful !

I hope that you can help me with my problem, since it is more or less specific..

I’m trying to calculate Cronbach’s alpha for a specific purpose: inter-rater reliability (rather than internal consistency reliability) where raters used Liker-scale to judge some behaviors shown in videos. 8 items were used for the Likert-scale. Until now it seems fine, I can measure Cronbach’s alpha coefficient for one behavior (one observed video) while considering 8 rows (for the 8 items) and n columns (for n raters).

However, there are many videos, and each rater evaluated only 16 videos (among around 600 videos). Each video was rated 24 times. My purpose is to measure the inter-rater reliability for each group of videos (the videos that share a common concept), and later, the overall inter-rater reliability..

My proposition is to measure the Cronbach’s alpha coefficient for each video, and then to measure the average of those coefficients in order to asses the global Cronbach’s alpha coefficient related to one group of videos. I only need to know whether this approach is reliable, I mean computing the average of n Cronbach’s alpha coefficients (that were computed separately for each video) to get the inter-rater reliability for a group of videos makes a sense !

Thank you in advance for your help, I really appreciate it since I could not find an answer anywhere..

Best regards,

Ness

Ness,

I am not sure what you mean by measuring Cronbach’s alpha for one video. I would think that you would need more than one video even to use Cronbach’s alpha, unless you are comparing what you call “items”. In your explanation you reference 8 items, but I am not sure what the items represent.

My key question to you is why you want to use Cronbach’s alpha in this way, when, if I understand what you are trying to accomplish well enough, there are other tests which fit better with your goal? Perhaps you should be using Fleiss’s Kappa instead. See, for example, the webpage http://www.real-statistics.com/reliability/fleiss-kappa/.

Charles

Pls can someone send me how to solve cronbach alpha of the example1 above.thanks

Hi Charles,

Thanks for your prompt response and the info that you provided. I’ll download the software and give it a try. The 15-day trial version of the SPSS that I initially used kept outputing -4 as my alpha and it was frustrating.

Let me try again and if I encounter any problem, I’ll let you know.

Thanks again.

Ricky

Hello Charles,

I really need your help.

I am process of Writing my thesis and I need to find Cronbach’s alpha to state the reliability of my research. I tried usingg SPSS but I keep getting -4 as the alpha which I know is not right. It is a likert scale based questioonaire with a scale of 1 to 5 and the number of questions that I selected to find the Cronbach’s alpha is 10.

I will be glad to send you the questionnaire to see how you can help me.

regards, ricky

I would be willing to send you the data that I used.

I will be very grateful if you can assist me to solve this problem.

Ricky,

The referenced webpage explains how to calculate Cronbach’s alpha in Excel. If you have downloaded the Real Statistics software you can use the formula =CRONALPHA(R1) where R1 is the range containing your data (formatted as described on the webpage) and the software will do all the calculations for you.

Charles

Hi Charles,

I think I got it. Is it the sum of all the variance?

I got 1 different last digit than yours, instead of 3, I got 4.

I don’t know if it matters or not or if I am mistaken.

Thx again.

Arin,

Yes. It is the sum of the variances. I used the formula =SUM(B17:L17) which sums the values in each of the cells B17 through L17. Although the value displayed for each of these cells contains 5 digits, more than this number of digits are actually stored. If you simply add the numbers that are displayed you get a number which ends in 4. If instead you add the complete numbers (as the formula =SUM(B17:L17) does) then the result will end in 3.

Charles

Hi Charles,

Thanks so much for the explanation. It’s really easy to follow.

However, I have a question about example 1. How do you get the value for cell B20?

In figure 2, you just said its from M17. But how do you calculate it?

Any explanation on this would be very much appreciated.

Thx

Arin,

M17 contains the formula =SUM(B17:L17).

Charles

Thanks so much indeed Charles. I got it.

You have a Blessed day.

Best,

Hi Charles,

Thanks so much for this site. Very helpful.

Reading from one of your observations above, and I quote, “If the variances of the xj vary widely, the xj can be standardized to obtain a standard deviation of 1 prior to calculating Cronbach’s alpha”, end of quote. just wondering how do you do that. Would be grateful for your response.

Thanks

Hi Luke,

If you have a column of data in range A1:A10. You calculate the mean of the values (say in cell A12) by =AVERAGE(A1:A10) and the standard deviation (say in cell A13) by =STDEV(A1:A10). Now put the formula =STANDARDIZE(A1,A$12,A$13) in cell B1. B1 now contains the standardized value of the data in cell A1. If you highlight the range B1:B10 and press Ctrl-D then column B will contain the standardized values of all the data in range A1:A10.

Charles

Hye, i would like to run cronbach alpha on each item as my on likert scale..as you show in example 4 that run crunbach alpha on all the items..

How can i do it on each item?

Possible?

Seems it does not clear with me here..

Thanks

I’m not sure what you mean by “run cronbach on each item”. You run cronbach on all the items. You can also run cronbach on all the items except one (as in Example 2). In any case, how you calculate cronbach is described on the referenced webpage and you can get more information by looking at the example worksheet which you can download for free at webpage http://www.real-statistics.com/free-download/real-statistics-examples-workbook/.

Charles

hi. i gor high Cronbach for my 4 variables but the correlations are zero. Is it somehing wrong wih my calculations or the data? Need help. Tq

Dee,

Can you send me a file with the example so that I can take a look at it?

Charles

Thank you very much for this website!! Absolutely excellent and very useful…!

Hi there – hopefully you can answer me

I would like to run a Cronbachs alfa test (or similar if you have an alteriour suggestion) to include it in my reliability part. (using SPSS)

I have a mixture of variables – Nominal, Ordinal and scale.

Which have a whole range of diferent ranges in values. Some variables have 0-1, others have 1-5, while others go from 0-14000. None are similiarly kategorized. Like, sex is a 0-1. Work hours range pr week from 0-30. A satisfactory question ranges from 1-5, rent payed ranges from 500- 13000… etc..

I’ve kategorized rent, and when I lump them all togheter in SPSS – Cronbachs alfa I get a Cronbachs Alfa of around ,500.

BUT – I have a feeling this is quite pointless? Because of the state of the values – is this so, is Cronbachs Alfa usless – how would I explain that I can’t use Cronbachs Alfa? My corriculum states that it is restricted to indexes – But I can’t quite understand what this means by “indexes”. Certainly we have an index of questions that all aim to answer an underlying question – but it’s not an index where all numbers are grouped simillarilly – so i am unsure what it means… Can’t I use it ?

Colene,

I have read a lot of conflicting information regarding the subject you are raising and so I don’t have a precise answer for you. My understanding is that Cronbach’s alpha is most relevant when the test is evaluating a single “factor”. You can certainly calculate Cronbach’s alpha even if the questionnaire contains a mix of multiple choice, true-false and other types of questions. What I would be especially cautious about is when the test is performing different types of evaluations (e.g. Likert scale to assess your satisfaction with a product plus multiple choice to assess your ability to use the product). You definitely need to calculate separate Cronbach alpha for each concept/factor that you are testing.

I’m nor sure what your curriculum means by “indexes”. Cronbach’s alpha itself can be used as an index, but it doesn’t sound like this is what is meant.

The following are a couple of articles on the web that may give you further information (although they may confuse things even more). I suggest that you speak to your professor to get further insights from him/her. Please share with the rest of us any insights you glean.

http://www.ctbassessments.com/assessment_insights/february_2011/research_insights.html

http://psychweb.psy.umt.edu/denis/datadecision/front/cortina_alpha.pdf

Charles

Regarding example number 3, I have performed Anova: Two Factor without Replication in Excel, but could not see Alpha value as it has been shown in Figure 5 above.

Zargoon,

I just rechecked and I believe it is correct. Please make sure you analyzed the data from Example 1 (0’s and 1’s only) and not from Example 2.

Charles

You have to type the alpha value (and that line input) in yourself if using Excel….

Sorry, but I don’t know what you are referring to.

Charles

Hello,

I am having trouble with measures that has items with reverse coding. When I calculate the data with reverse coding, the Cronbach’s Alpha is very low (.2), however when I calculate the Cronbach’s Alpha without using the reverse coding, it is very high (.9). I am wondering if you know why that happens, and what I should do?

Dianna,

It is hard for me to tell without seeing the data, but I can think of the following possibilities:

– the questions that you have identified as reverse coded are not really reverse coded

– there is an error in the coding or calculation

– reliability is low (not sure why alpha would be high if reverse coding is not done)

Charles

Hello,

I am using an instrument that has been previously tested for reliability with a Cronbach Alpha of .91 and .93 for overall instrument (there are two variables being tested). The instrument has four subscales and each subscale has a Cronbach Alpha number of >.61 (three subscales are in the .80 range). I used the same instrument and tested Cronbach Alpha and received an overall number of .87 and .86. All of my subscales are substantially lower than the original testing by the instrument author. How do I explain that? We used a similar population. Her study was larger with 64 participants and mine had 40.

Thank you.

Colleen

Colleen,

A few reasons why your alpha values may differ are: population not really the same as that for the original instrument, randomness (you say that your values were substantially lower, but it may turn out that the difference is not statistically significant), differences in the way the test was administrated(e.g. a noisy environment, people under stress, etc.).

Charles

how did you get the var in example 4. i just don’t get it how it can be computed. is there a formula needed?

Krisliz,

Cell B24 contains the formula =VARP(L4:L18).

Charles

Hi,

I’m using a sample of 40 and I have 2 sets of questions, one with 9 and the other with 10 questions. I’ve calculated the alpha for both using the formula as you explained and then using the Real Statistics toolpack. I’m getting different alphas when I use the two methods with the toolpack function giving an alpha higher by around 0.08. Why is that?

Karen,

Please email me the Excel worksheet(s) so that I can figure out happened.

Charles

I’m not a student of mathematics. I am leading a research in applied linguistics and I need to calculate a result using Cronbach alpha. I asked my students about the extent to which they became autonomous after introducing language learning strategies on a five-graded Likert scale with scores from 1 to 5. However, I didn’t know how to calculate. On the vertical column, I have to mention the respondents (students) and on the horizontal line, I have only one question (so one item??). Could I calculate alpha with only one column?

I am seeking guidance and I would be very grateful if you could help me.

Thank you.

Hadia,

You can’t measure consistency between the items (which is what Cronbach’s alpha does) since you only have one item. You need more than one item to use Cronbach’s alpha. My question to you is why do you want to use Cronbach’s alpha? What are you trying to demonstrate?

Charles

Enriching information.Kudos

What percentage of the population is recommended for purpose of determining Cronbach Alpha?

Joseph,

I have generally used the entire population of subjects who have filled in a questionnaire to calculate Cronbach’s alpha.

Charles

Do you have a reference for this”A commonly-accepted rule of thumb is that an alpha of 0.6-0.7 indicates acceptable reliability”? Thank you very much!

Vien,

I believe that I got this from the following reference (this is also referenced in Wikipedia)

Ref 1: Kline, P. (2000). The handbook of psychological testing (2nd ed.). London: Routledge, page 13

I have generally seen that .7 is viewed as the minimum acceptable level. Here is such a reference.

Ref 2: George, D., & Mallery, P. (2003). SPSS for Windows step by step: A simple guide and

reference. 11.0 update (4th ed.). Boston: Allyn & Bacon. p. 231

The rules of thumb there are:

> .9 – Excellent, _ > .8 – Good, _ > .7 – Acceptable, _ > .6 – Questionable, _ > .5 – Poor, and < .5 – Unacceptable

Here is yet another example where .6 is used as the minimal acceptable level.

Ref 3: "Cronbach’s alpha (Cronbach, 1951) which quantifies the degree of internal consistency (reliability) of a set of items, was calculated for each subscale, as well as the overall scale. In general, a Cronbach’s alpha of at least .7 is the criterion used to establish an acceptable level of reliability. However, the recommended minimum Cronbach’s alpha for exploratory studies is .6" (Nunnally, J.C. (1978). Psychometric Theory (2nd ed.). New York: McGraw Hill; Robinson, J. P., Shaver, P. R., & Wrightsman, L. S. (Eds.). (1991). Measures of personality an social psychological attitudes. San Diego: Academic Press).

Thank you Charles. Your response is highly appreciated. My apology for this late reply. God bless.

If I have a 2 group study and want to conduct a cronbach alpha test, do I have to separate it into 2 groups?

Mabel,

It really depends on what you want to study. E.g. if the groups are just a random split of the sample you could use the split-half method instead of Cronbach’s alpha to measure reliability.

Charles

Great site. Thank you.

I am looking for simple consistencies within a singular column of measured data, to see how volatile it is. I tried to use your CRONALPHA function to simply analyze one column of data of designated cells, and am getting the “#VALUE” response.

Am I missing something? Thanks much.

Jonathan

Jonathan,

Cronbach’s alpha is not designed to do what you want. There is no internal consistency to measure when only have one response per subject. To use CRONALPHA you need at least two columns.

Charles

Thank you Charles.

Perhaps I am asking the wrong question. Let’s say I have 120 responses from one subject over time, and what I am trying to do is to look at the consistency and repeatability of the answers from that subject. In other words, I want to see if the data, once I reach a certain point, has a stability to it. While I could do an even-odd correlation from the set, 1st-half / 2nd-half, or something like that, that introduces the limitations you discuss.

Is there something that considers the stability of a great number of data from one subject alone, to look at stability purposes? Thanks very much for a great site.

Jonathan,

I am afraid that I don’t have much help to offer. I checked the Internet and found a number of sites addressing

statistical analysis of single subject dataandsingle subject consistency statistics, including the following:http://www.physther.net/content/74/8/768.full.pdf

http://www.u.arizona.edu/~dusana/psych290Bpresession06/notes/Ch%2014%20Single-Subejct%20Research%20Designs.ppt

Unfortunately these sites seem to be addressing something different from what I think you are looking for.

Charles

Hi,

I have approximately 200 respondense, could you please let me know how to ho about getting the results of the Cronbach test. Do I have to enter a figure for each respondent to a question?

Thanks,

MC.

Mukesh,

If you have say 200 people taking the test and each test has 20 questions, then you would create a range similar to that in Figure 1 with 200 rows and 20 columns. You need to enter a figure to each question.

Charles

Hi,

Thank you for the great explanation.

My work includes using a 5 point scale by 2 separate groups of raters (trained Vs untrained in that field). Each group will rate 20 different cases.

My questions are:

1- Do I need to use Cronbach’s alpha for each case separately (20 cases x 2 groups) (i.e. 40 times)?

2- How can I assess the consistency within each group for all the cases collectively?

3- How can I know if there is a significant difference between the 2 groups for all the cases collectively?

Apologies for the long questions, but I really appreciate your help.

Best regards,

Dalia

Hi Dalla,

From what you have described I have the following suggestions:

1. to assess the agreement/disagreement between the two raters you probably want to use something like Cohen’s kappa

2. to assess the consistency within each of the two groups you can use Cronbach’s alpha

3. to assess whether there is a significant difference between the two groups you can use either a t test or Mann-Whitney

All of these tests are described on the Real Statistics website and the Real Statistics Resource Pack can be used to carry out each of these tests.

Charles

this method was very helpful for an average student like me,thank you very much…

Hi. I got negative cronbach alpha which made it unacceptable but the questions are just right and fit to my study. What should I do? Thank you for your response. God bless.

Hi Krsna,

There are an number reasons for getting a low (or even negative) value for cronbach’s alpha for a perfectly valid test. Two reasons are reverse coding and multiple factors.

Reverse coding: Suppose you use a Likert scale of 1 to 7 with 1 meaning strongly disagree and 7 meaning strongly agree. Suppose two of your questions are: Q1: I like pizza and Q20: I dislike pizza. These questions ask the same thing, but with reverse scoring. In order to apply Cronbach’s alpha properly you need to reverse the scoring of the negatively phrased question, Q20. Thus if a response to Q20 is say 2, it needs to be scored as 6 instead of 2 (i.e. 8 minus the score).

Multiple factors: Cronbach’s alpha is useful where all the questions are testing more or less the same thing, called a “factor”. If there are multiple factors then you need to determine which questions are testing which factors. If say there are 3 factors, then you need to split the questionnaire/test into three tests, one containing the questions testing factor 1, one with the questions testing factor 2 and the third with questions testing factor 3. You then calculate cronbach’s alpha for each of the three tests. The process of determining these “hidden” factors and splitting the test by factor is called Factor Analysis. See the webpage http://www.real-statistics.com/multivariate-statistics/factor-analysis/ for more details about how to do this in Excel using Real Statistics.

Charles

Thank you, Sir. It´s a great example. it contains all that I have searched. ¿Aren´t you teacher?

how can i calculate five point scale survey in excel and get the cronbach alpha??? its quite confusing.

Mary Ann,

Enter your data as in Example 4 on http://www.real-statistics.com/reliability/cronbachs-alpha/. The example shows what to do for a seven point scale, but a five point scale works exactly the same. Then carry out the calculations as in Figure 6 on the same webpage or simply use CRONALPHA(R1) where R1 is the the range containing your data (without headings). CRONALPHA is a supplemental function which is contained in the Real Statistics Resource Pack.

Charles

the test contains 18 questions in which 12 mcq’s,5 two marks questions and 1 three marks questions ,how can i calculate alpha coefficient

Hi Jeena,

I believe that for each question you simply score 1 for a correct answer and 0 for an incorrect answer, whether the multiple choice question has 2,3 or 4 choices. Then follow the procedure described in Example 1 or Example 3 on webpage http://www.real-statistics.com/reliability/cronbachs-alpha/. You can also use the supplemental formula CRONALPHA provided in the Real Statistics Resource Pack.

Charles

Dear Charles,

as if you have different parts of questionnaire e.g. Yes, No and with 5 scale options.. So would there be two separate evaluations of C Alpha or may i find the one single value as well for both segregated portions of questionnaire ?

Dear Hammad,

Are you saying that if they answer Yes they have to evaluate something on the scale of 1 to 5, while if they say No they have a separate evaluation also using a scale of 1 to 5?

Charles

Hi,

I would like to know, if i have 5 variables that i want to categorize into a single variable. Out of them, 4 questions are likert scale and 1 is a yes/no question. how can i find the chronbach alpha for the 5 questions?

Hi Koel,

Provided all the questions measure the same thing you can calculate one Cronbach’s alpha for all the questions.

However, I would not mix Likert scale questions (rate your opinion of Columbus from 1 to 5) with knowledge questions (did Columbus discover America? T/F).

Generally for the Likert scale questions use the Likert scale (say 1 to 5) in calculating Cronbach’s alpha. For True/False questions you can use values that correspond to the Likert scale (e.g. 5 if True means you have an extremely high opinion of Columbus and 3 if No means that you have neutral opinion). Most of the time True/False questions are measuring knowledge and so as stated above it is not appropriate to mix them with Likert scale questions in calculating Cronbach’s alpha.

In general with knowledge-based questions (e.g. True/False or multiple choice questions) you can code them as 1 for the correct answer and 0 for an incorrect answer when calculating Cronbach’s alpha.

Charles

Charles