Welcome to the *Real Statistics Using Excel *website. As described on the Home page, this website provides the resources necessary to perform statistical analysis in the Excel environment.

In order to make this website more valuable to the users, we welcome your comments, suggestions and feedback.

We would especially like to hear about any errors in the website or factual errors in the statistical analyses presented. Please also let us know about any bugs you find in the supplemental functions or data analysis tools in the Real Statistics Resource Pack or Examples Workbook.

We also welcome your suggestions about how to improve the site or additional topics we should include in the site.

If you have already downloaded the Real Statistics Resource Pack, thanks for choosing to do so; we hope that you will return to this website for additional information and for updates to the software.

Charles Zaiontz

Hi,

I was thrilled to see that you had a test for normality (Shapiro-Wills). Concerned that my data series might not be normally distributed, I put in the range of my values (stock returns) — 5000 of them!

However, it appears that the Shapiro functions are limited to ranges of less than 50 cells.

Any idea of what I might be doing wrong?

Thanks,

Craig

Hi Craig,

Yes, you are correct: Shapiro-Wilk is limited to 50 data elements. The idea of the test is to determine whether a random sample is normally distributed. By the Central Limit Theorem, random samples with over 50 elements tend to be pretty normally distributed and so the test doesn’t need to go beyond about 50 elements.

However, if your data is not coming from a traditional random sample, then it may not be normally distributed. To check whether it is normally distributed you can use a QQ plot, which is also available in the Real Statistics Resource Pack. If the output looks reasonably like a straight line then the data is normally distributed. You can also use a chi-square Goodness of fit test on the tabular output from the QQ plot generated by the Real Statistics Resource Pack.

Depending on how you will be using the data it often sufficient to test whether the data is reasonably symmetric. In this case a Box Plot is a reasonable test.

Please let me know if you need further help.

Charles