# Dichotomised data-how to test for a difference(potentially independent t-test) SPSS

I am using a 2X3X4 design, which has the following ivs: Social Anxiety Scores(formed two groups,essentially high and low scores) 3 emotions and 4 gaze directions. I need to test if the groups are essentially different(before I do my mixed anova). I thought I could do this through an independent t-test. As I have tried to do one, it has tested the differences between each condition individually. How can I test them for difference, is it a matter of how I present the data in the data viewer? Or is the another method to do it?

• You talk about your IVs but not your DV (though the title suggests it takes two values). This would suggest t-tests are not suitable, but why is it important to test for univariate differences before doing an ANOVA? Wouldn't Simpson's paradox suggest that's potentially miisleading? – Glen_b Mar 13 '14 at 20:35
• Sorry, I didn't make myself as clear as I hoped, I thought I would overcomplicate it with too much information. My DV is the proportion individuals feel looked at. With regards to the independent T-test, because I have decided how I am grouping the social anxiety scores myself, and am not using an "official" boundary to split them, I need to see if there is a true difference between these groups first before considering my DV/ANOVA. I am not combining the data so I do not see how Simpson's paradox applies. – user41877 Mar 15 '14 at 12:25
• (1) What are you testing the groups for a difference on? (2) When you say the DV is "the proportion individuals feel looked at" do you mean the fraction of time for each individual or the count of the people who feel looked at divided by the total number of people? (Either way, you're very likely to have heteroskedasticity, and ANOVA may not be appropriate for data bounded between 0 and 1). – Glen_b Mar 15 '14 at 19:50
• I have managed to do what I meant to do now. Thank you for your input. – user41877 Mar 19 '14 at 16:25