I have two categorical variables - IV = "gender" (M, F) and DV = "Who was guilty" ("man", "woman", "neither", "both"). Participants viewed a scene involving a dispute between a man and woman. I have carried out a chi square test of correlation between gender and who was guilty, and no overall correlation was found. However, I notice from the chart that a relatively much higher number of women thought that the woman was guilty. The problem is, chi square does not give me a coefficient for each level of guilt (man, woman, neither, both) separately but only a general one for the DV as a whole. What test can I use if I want to look more closely at the relation between gender and one of the levels in my DV? In effect I want to look purely at the level "woman" as being the guilty party and relation to gender, i.e. to see if the difference between the number of men who thought the woman was guilty and the number of women who thought the woman was guilty is significant. Or is there any way of performing the chi square test while filtering out the other levels in the analysis? I did try to filter out the 3 other levels ("man", "neither", "both") through "Select cases" function, but my out put then said "No statistics were computed as "who was guilty" is a constant". I thought about doing a one way Anova as the post-hoc test does give info about each individual level, but I have two nominal variables while I think Anova requires at least one interval level variable. Any guidance would be appreciated.