Possible to test correlation between 2 nominal scaled questions?

#1
Hey

We have one more question, here in the middle of our Bachelor thesis.

We have two questions in our questionnaire, both are nominal scaled, and they don't follow a normal distribution. Is it possible to test for any correlation between the level of education in q1 and q2? We have done a crosstab to see the difference, but is it possible to test it?

1. What is your level of education?
No schooling completed
High School Graduate
...
Masters degree

2. What is your mothers highest level of education?
The possible answers are the same as in question 1.

Thanks

Kasper and Claus
 
#2
Hi Kaspar and Claus,

Yes. If you're doing the Crosstab in Descriptives, click on the Statistics button. You can choose either Chi-square or Fisher's exact. Both will give you p-values.

The Chi-square has stricter assumptions, so if you get a little warning under the table output about your expected cell sizes being smaller than 5, you need to use Fisher's exact.

Otherwise, the Chi-square is the more common statistic.

Karen
 

CB

Super Moderator
#3
How about just looking at the Spearman's rho or Kendall's tau correlation coefficients? (Analyse / Correlation). These are both rank-based non-parametric correlation measures that don't require particular distributional assumptions.

It looks like your variables are ordinal rather than nominal btw - doesn't each level of response represent a higher level of education?