In this summary you will learn how distinguish parametric tests from non-parametric tests and how to conduct non-parametric tests using SPSS. Click here for more information on parametric tests. 

In this table you can see what test to use:

When performing a test, you first have to test your data for normality. You do this by following these steps in SPSS:

Then you get two tests of normality. You use the Smirnov when using >2000 participants, otherwise you use the Shapiro. If this test is significant, the assumption of normality is NOT met. 

If the assumption of normality of not met, you can still conduct an analysis, but you have to test with non-parametric tests. In the table above you can see what test to use. 

Non-parametric vs parametric:

• Pro: non-parametric is robust against the violation of the normality assumption
• Con: less powerful

You can also log-transform. Non-parametric vs. los-transform:

• Pro: non-parametric outcome is easier to interpret than log-transform
• Con: less powerful

Mann-Whitney U-test

Look at the exact sig. (2-tailed) for the significance. If it is significant, calculate the effect size. 

Kruskall-Wallis test

As output you get a decision. You also get a subset table. If the variables are in different subsets, they differ signifantly from eachother. You do not get a seperate p-value for this. When reporting, also report the H and the freedom of degrees.

Wilcoxon Signed-Rak test

Report at least the Z-value, the p-value and the effect size. 

Friedman test

How to calculate effect sizes

For an ANOVA, you can just give the partial eta-squared. For others you might have to calculate it yourself using:

The effect sizes can be interpreted using this table (also report the interpretation!):

Questions? Let me know in the contribution section!

Follow me for more summaries on statistics!