Hi everyone!

I run a questionnaire which had 4 conditions, participants entering the questionnaire were randomly assigned to one of them. In each different condition they had to choose between two t-shirts. I want to test if there is a correlation between the conditions and their T-shirt choice. Being all conditions categorical and binary outcomes, I supposed I have to run a chi-square test to test it. 

Before I had 4 variables for each condition, but I couldn't run the test, because I always got a warning that the "crosstabulations were empty" (as each participant was assigned to a different condition the result was that there was data in separate rows for each condition and there was no overlap in responses between conditions).

So, I have created a variable called "condition type" (from 1 to 4) and a variable called "t-shirt type" and now I suppose I have to run a Chi Square test of independence between these 2 variables, but my question is, do I have to run a test for each condition separately, or is one test between "condition type" and "t-shirt type" enough? If I have to test them separately, how can I do it, having a single variable (condition type) containing all 4 conditions?

Thank you very much!

Hi Elisabetta

If you want simply to test the independence between the variables "condition type" and "t-shirt type" , then it is enough to proceed with Chi Square test of independence. The contingency table (or cross tabulation or cross tab) consists of 4 rows of different conditions and two columns for t-shirt. Or you can choose t-shirt row wise and the conditions column wise. In SPSS please follow Analyse →Descriptive Statistics →Cross tab. Then choose rows and columns, click Statistics →Chi Square →Ok.

Hi! Thank you very much! As I am quite new to statistics and to SPSS, do you think I could test also something else (despite the independence), having these two variables? 

Because apart from that I was thinking about doing a regression analysis to see if education and income have an impact on the t-shirt type choice.

Hi

You can try with regression but be careful while choosing the type of regression. For instance, if you want to predict a numerical quantity, then proceed with linear (simple or multiple) regression. Similarly, if you want to predict a category, for example, which type of t-shirt is preferred, in those cases, apply logistic regression. Other types of regressions such as Poisson regression, Polynomial regression, etc are also there in the literature.