We consider a random variable *x* and a data set *S = *{*x*_{1}*, x*_{2}*, …, x _{n}*} of size

*n*which contains possible values of

*x*. The data set can represent either the population being studied or a sample drawn from the population. The mean is the statistic used most often to characterize the center of the data in

*S*. We now consider the following commonly used measures of variability of the data around the mean, namely the

**standard deviation, variance**,

**squared deviation**and

**average absolute deviation**.

In addition we also explore three other measures of variability that are not linked to the mean, namely the **median absolute deviation**,** range** and **inter-quartile range**.

Of these statistics the variance and standard deviation are most commonly employed.

**Excel Functions**: If R is an Excel range which contains the data elements in *S* then the Excel function which calculates each of these statistics is shown in Figure 1. Functions marked with an asterisk are supplemental functions found in the Real Statistics Resource Pack, although equivalent formulas in standard Excel are described later.

Statistic |
Excel 2007 |
Excel 2010/2013 |
Symbol |

Population Variance | VARP(R) | VAR.P(R) | σ^{2} |

Sample Variance | VAR(R) | VAR.S(R) | s^{2} |

Population Standard Deviation | STDEVP(R) | STDEV.P(R) | σ |

Sample Standard Deviation | STDEV(R) | STDEV.S(R) | s |

Squared Deviation | DEVSQ(R) | DEVSQ(R) | SS |

Average Absolute Deviation | AVEDEV(R) | AVEDEV(R) | AAD |

Median Absolute Deviation | MAD(R) * | MAD(R) * | MAD |

Range | RNG(R) * | RNG(R) * | |

Inter-quartile Range | IQR(R, b) * |
IQR(R, b) * |
IQR |

**Figure 1 – Measures of Variability**

**Observation**: These functions ignore any empty or non-numeric cells.

**Variance**

**Definition 1**: The variance is a measure of the dispersion of the data around the mean. Where *S* represents a population the **population variance** (symbol ** σ^{2}**) is calculated from the population mean

*µ*as follows:

Where *S* represents a sample the **sample variance** (symbol ** s^{2}**) is calculated from the sample mean

*x̄*as follows:

The reason the expression for the population variance involves division by *n* while that of the sample variance involves division by *n* – 1 is explained in Property 3 of Estimators, where division by *n* – 1 is required to obtained an unbiased estimator of the population variance.

**Excel Function**: The sample variance is calculated in Excel using the worksheet function **VAR**. The population variance is calculated in Excel using the function **VARP**. In Excel 2010/2013 the alternative forms of these functions are **VAR.S** and **VAR.P**.

**Example 1**: If *S* = {2, 5, -1, 3, 4, 5, 0, 2} represents a population, then the variance = 4.25.

This is calculated as follows. First, the mean = (2+5-1+3+4+5+0+2)/8 = 2.5, and so the squared deviation *SS* = (2–2.5)^{2} + (5–2.5)^{2} + (-1–2.5)^{2 }+ (3–2.5)^{2} + (4–2.5)^{2} + (5–2.5)^{2} + (0–2.5)^{2 }+ (2–2.5)^{2} = 34. Thus the variance = *SS*/*n* = 34/8 = 4.25

If instead *S* represents a sample, then the mean is still 2.5, but the variance = *SS*/(*n–*1) = 34/7 = 4.86.

These can be calculated in Excel by the formulas VARP(B3;B10) and VAR(B3:B10), as shown in Figure 2.

**Figure 2 – Examples of measures of variability**

**Observation**: When data is expressed in the form of frequency tables then the following properties are useful. Click here for the proofs of these properties.

**Property 1**: If *x̄* is the mean of the sample *S = *{*x*_{1}*, x*_{2}*, …, x _{n}*}, then the sample variance can be expressed by

**Property 2**: If *µ* is the mean of the population *S = *{*x*_{1}*, x*_{2}*, …, x _{n}*}, then the population variance can be expressed by

**Standard Deviation**

**Definition 2**: The **standard deviation** is the square root of the variance. Thus the population and sample standard deviations are calculated respectively as follows:

** Excel Function**: The sample standard deviation is calculated in Excel using the worksheet function

**STDEV**. The population standard deviation is calculated in Excel using the function

**STDEVP**. In Excel 2010/2013 the alternative forms of these functions are

**STDEV.S**and

**STDEV.P**.

**Example 2**: If *S* = {2, 5, -1, 3, 4, 5, 0, 2} is a population, then the standard deviation = square root of the population variance = = 2.06

If *S* is a sample, then the sample standard deviation = square root of the sample variance = = 2.20

These are the results of the formulas STDEVP(B3:B10) and STDEV(B3:B10), as shown in Figure 2.

**Real Statistics Functions**: The Real Statistics Resource Pack furnishes the following array functions:

**VARCOL**(R1) = a row range which contains the sample standard variances of each of the columns in R1

**STDEVCOL**(R1) = a row range which contains the sample standard deviations of each of the columns in R1

**VARROW**(R1) = a column range which contains the sample standard variances of each of the rows in R1

**STDEVROW**(R1) = a column range which contains the sample standard deviations of each of the rows in R1

**Example 3**: Use the VARCOL and STDEVCOL functions to calculate the sample variance and standard deviation of each of the columns in the range L4:N11 of Figure 3.

The formula =VARCOL(J4:L11) produces the first result (in range J15:L15), while the formula =STDEVCOL(J4:L11) produces the second result (in range J16:L16). Remember that after entering either of these formulas you must press **Ctrl-Shft-Enter**.

**Figure 3 – Sample Variance and Standard Deviation by Column**

**Property 3**: If the population {*x _{1}, x*

_{2}

*, …, x*} has mean

_{n}*µ*and standard deviation

_{x}*σ*and the population {y

_{x}_{1}, y

_{2}, …, y

*} has mean*

_{m}*µ*

_{y }and standard deviation

*σ*

_{y}, then the variance of the combined population is

Thus if *µ _{x}* =

*µ*

_{y }the combined population variance would be

**Property 4**: If the sample {*x*_{1}*, x*_{2}*, …, x _{n}*} has mean

*x̄*and standard deviation

*s*

_{x}and the sample {y

_{1}, y

_{2}, …, y

*} has mean*

_{m}*ȳ*and standard deviation

*s*

_{y}, then the variance of the combined sample is

Thus if *x̄ = ȳ*, the combined sample variance would be

**Example 4**: Find the mean and variance of the sample which results from combining the two samples {3, 4, 6, 7} and {6, 1, 5}.

**Figure 4 – Calculation of combined mean and standard deviation**

The data in the two samples is given in the range B3:C7 of Figure 4. From these, the mean, variance and standard deviation are calculated for each of the two samples (ranges B12:B15 and C12:C15). Using Property 4, we can calculate the mean and variance of the combined sample (D13 and D14).

If we simply combine the two samples we obtain the data in the range F3:F10, from which we can calculate the mean, variance and standard deviation in the normal way (range D12:D18). As we can see the results are the same.

**Squared Deviation**

**Definition 3**: The **squared deviation** (symbol ** SS** for

**sum of squares**) is most often used in

**ANOVA**and related tests. It is calculated as

** Excel Function**: The squared deviation is calculated in Excel using the worksheet function

**DEVSQ**.

**Example 5**: If *S* = {2, 5, -1, 3, 4, 5, 0, 2}, the squared deviation = 34. This is the same as the result of the formula DEVSQ(B3:B10) as shown in Figure 2.

**Average Absolute Deviation**

**Definition 4**: The **average absolute deviation** (**AAD**) of data set *S* is calculated as

**Excel Function**: The average absolute deviation is calculated in Excel using the worksheet function

**AVEDEV**.

**Example 6**: If *S* = {2, 5, -1, 3, 4, 5, 0, 2}, the average absolute deviation = 1.75. This is the same as the result of the formula AVEDEV(B3:B10) as shown in Figure 2.

**Median Absolute Deviation**

**Definition 5**: The **median absolute deviation** (**MAD**) of data set *S* is calculated as

Median {|*x _{i}* – | :

*x*in

_{i}*S*}

where = median of the data elements in *S*.

**Excel Formula**: If R is a range which contains the data elements in *S* then the MAD of *S* can be calculated in Excel by the array formula:

=MEDIAN(ABS(R-MEDIAN(R)))

Even though the value is presented in a single cell it is essential that you press **Ctrl-Shft-Enter** to obtain the array value, otherwise the result won’t come out correctly. This function only works properly when R doesn’t contain any empty cell or cell with a non-numeric value.

Alternatively, you can use the supplemental function **MAD**(R) which is contained in the Real Statistics Resource Pack. This function works properly even when R contains empty cells and/or cells with non-numeric values. You don’t need to press **Ctrl-Shft-Enter** to use this function.

**Example 7**: If *S* = {2, 5, -1, 3, 4, 5, 0, 2}, the median absolute deviation = 2 since *S = *{-1, 0, 2, 2, 3, 4, 5, 5}, and so the median of *S* = (2+3)/2 = 2.5. Thus MAD = the median of {3.5, 2.5, 0.5, 0.5, 0.5, 1.5, 2.5, 2.5} = {0.5, 0.5, 0.5, 1.5, 2.5, 2.5, 2.5, 3.5}, i.e. (1.5+2.5)/2 = 2.

You can achieve the same result using the supplemental formula =MAD(E3:E10) as shown in Figure 2.

**Observation**: This metric is less affected by extremes in the tails because the data in the tails have less influence on the calculation of the median than they do on the mean.

**Range**

**Definition 6**: The **range** of a data set *S* is a crude measure of variability and consists simply of the difference between the largest and smallest values in *S*.

**Excel Formula**: If R is a range which contains the data elements in *S* then the range of *S* can be calculated in Excel by the formula:

=MAX(R) – MIN(R)

Alternatively, you can use the supplemental function **RNG**(R) which is contained in the Real Statistics Resource Pack.

**Example 8**: If *S* = {2, 5, -1, 3, 4, 5, 0, 2}, the range = 5 – (-1) = 6. You can achieve the same result using the supplemental formula =RNG(E3:E10) as shown in Figure 2.

**Inter-quartile Range**

**Definition 7**: The** inter-quartile range** (**IQR**) of a data set *S* is calculated as the 75% percentile of *S* minus the 25% percentile. The IQR provides a rough approximation of the variability near the center of the data in *S.*

**Excel Formula**: If R is a range which contains the data elements in *S* then the IQR of *S* can be calculated in Excel by the formula:

=QUARTILE(R, 3) – QUARTILE(R, 1)

In Excel 2010/2013 there is a new version of the quartile function called QUARTILE.EXC. An alternative version of IQR is therefore

=QUARTILE.EXC(R, 3) – QUARTILE.EXC(R, 1)

See Ranking Functions in Excel for further information about the QUARTILE and QUARTILE.EXC functions. Alternatively, you can calculate the inter-quartile range via the supplemental function **IQR**(R, *b*) which is contained in the Real Statistics Resource Pack. When *b* = FALSE (default), the first version of IQR is returned, while when *b* = TRUE the second version is returned.

**Example 9**: If *S* = {2, 5, -1, 3, 4, 5, 0, 2}, then the first version of IQR = 4.25 – 1.5 = 2.75, while the second version is IQR = 4.75 – 0.5 = 4.25. You can achieve the same result using the supplemental formulas =IQR(B3:B10) and =IQR(B3:B10,TRUE), as shown in Figure 2.

**Observation**: The variance, standard deviation, average absolute deviation and median absolute deviation measure both the variability near the center and the variability in the tails of the distribution which represents the data. The average absolute deviation and median absolute deviation do not give undue weight to the tails. On the other hand, the range only uses the two most extreme points and the interquartile range only uses the middle portion of the data.

Dear Mr. Zaiontz,

I’m looking for a method to evaluate the variability of some kinetic parameters of gait and found MAD more appropriate that coefficient of variation. My question is why you didn’t have used in the MAD formula the multiplier “b” as in the reference http://web.ipac.caltech.edu/staff/fmasci/home/statistics_refs/BetterThanMAD.pdf

[ Alternatives to the Median Absolute Deviation]

Thank you for your time,

Respectfully,

Daniel

Dear Daniel,

The standard calculation for the MAD function is the one that I have used. The b multiplier depends on the distribution. For large samples which are normally distributed, b * MAD is approximately equal to the standard deviation where b = 1.4826. For other distributions b would have a different value.

Charles

Dear Dr. Zaiontz,

My gratitude for this great support. I would like to know how to calculate common standard deviation from ttest paired. I need to get it to assess continuous data from crossover trial.

Thanks you very much.

Arturo

Arturo,

The formula for the standard deviation is the one given on the webpage http://www.real-statistics.com/descriptive-statistics/measures-variability/. Or you can use the Excel formula STDEV as usual. In the case of the paired t test you calculate the standard deviation on the difference between the data values for the two samples (as shown in Figure 2 of http://www.real-statistics.com/students-t-distribution/paired-sample-t-test/). The standard error is then this value divided by the square root of the sample size, as explained right after Figure 2.

If you download the Real Statistics Examples Workbook you can access the spreadsheets for all the examples on the website. From these spreadsheets you can see how all the calculations are done.

Charles