Statistics – nonparametric pairwise test

Last Monday 13/8 I studied a difference between “stress after” and “stress before”, to see if this difference was normal or skewed distributed.

The mean value of the difference was -0.29 and the median of the difference was -0.44.

So when it’s not normal distributed it’s more common to use the median value (very common when study salaries). If you want to use a statistics test and you don’t have normal distribution it’s common to use a nonparametric test.

The command is found here, and I choose related samples:

Note: you can also use the same type of tests in the “Legacy Dialogs”, there you find the older versions of the tests.

Next you click on the tab: fields and choose your paired variables. Here I put them in time order like this, because it’s easier to understand the descriptive statistics later:

The result tell me that the difference is statistically significant as the significance value is 0.037 (<0.05) and is also marked with a yellow color:

You have small popup menus in the bottom of the 2 windows, so you can switch between different results. But if you have a look of the right window, you can see the descriptive statistics:

Above you get information that the test is comparing “stress_after” minus “stress_before” with Wilcoxon Signed Rank test (se the bottom left part). The “Test statistics” is -2.086 and tell us that there has been a significant decrease of the stress. You also have a histogram of the differences that shows us that 16 of all differences are negative and only 9 are positive (:increased stress).

If you change the View in the bottom part of the window you can choose histogram over the original variables “Stress_before” and “Stress_after”, here below you see a description of “stress_after”.

Many researcher has to write a report of the results and when I worked with medical researchers, many wrote like this:

There has been a significant decrease of the stress, (p=0.037). The median decrease has been -0.44.

If you think that choices of methods is difficult, just talk to your contact person or send me an email and we can see if you would like to invest in a webtraining  or some consulting hours. My email: gunilla.rudander@crayon.com

Thanks for today, Gunilla Rudander

error: Content is protected !!