SPSS nonparametric tests are mostly used when **assumptions aren't met** for other tests such as ANOVA or t tests.Strictly, most “nonparametric tests” in SPSS are distribution free tests. Oddly, these two concepts are entirely different but often used interchangeably. Second, nonparametric tests are suitable for ordinal variables too. In rare cases they may have more statistical power than standard tests.

Not sure **which (nonparametric) test to use**? You'll quickly find the answer in Simple Overview Statistical Comparison Tests.

## This tutorial has 3 comments

## By Ruben Geert van den Berg on September 24th, 2018

Hi Loo!

Post hoc tests examine

whichproportions differ if the main chi-square independence test indicates that 2 variables are associated.For a 2 by 2 table, however, you're comparing only 2 proportions. So there's no question

whichproportions differ -there's only 2 of them.On top of that, a z-test for 2 independent proportions is preferred over the chi-square test here. The significance levels are identical but the z-test provides a confidence interval of the difference too -which is very informative. For some stupid reason, it's missing from SPSS but we built a simple tool for it.

You can also inspect the (Pearson) correlation between the 2 variables. For some dumb reason it's called a phi coefficient for 2 dichotomous variables but it's technically identical to a Pearson correlation.

Last, Bonferroni corrected pairwise z-tests for larger (2 by 3, 3 by 3...) tables are often preferred over the adjusted standardized residuals. You can request the latter by adding BPROP to the CELLS subcommand as in

CROSSTABS education BY marital_status

/CELLS COUNT COLUMN BPROP

/STATISTICS CHISQ PHI.

Note that these tests only compare column (not row!) proportions.

Hope that helps!

SPSS tutorials

## By Loo on September 24th, 2018

Hi Ruben, Post hoc test is available for chi square test of independence by using standard residual method, for example for a 3 by 3 chi square table. But, is it possible to do it for a 2 by 2 table?

## By Monalisa on September 21st, 2018

Great help!