Hello Hadeel,

Try ordinal regression, using 3 IVs: Group (2 levels, such as 0/1), Gender (2 levels, such as 0/1) and Group X Gender (just multiply group value by gender value), and your ordinal DV is the dependent variable.

This will permit you to determine which, if any, of the IVs is helpful in distinguishing among score categories on the DV.

Good luck with your work.

Following the answer of David Morse, you can try logistic regression. For example, by PROC LOGISTIC of SAS.

David Morse Thank for the reply. Could you please explain, Why we need to create a 3rd IV? (Group*Gender)? I ran the ordinal regression two times one with 2 IVs and another with 3 IVs as you recommend. There is a difference in the significance level but I didn't understand why?

Seyed Abbas Rafat I tried it it was almost the same results as ordinal regression. Thanks for the suggestion

Hello again Hadeel,

Yes, there very well might be a difference in the significance level, both for the overall model and for the "main effects" of gender and/or group. Why? In part because these are two different models. As well, because your first model (with gender and group as the lone IVs) presumes zero interaction between the two as regards score values on your DV.

The reason for the interaction term is that it allows you to ascertain whether specific combinations of gender and group behave differently from what you would predict by the value/level for gender and the value/level for group.

The presence of interaction signals that there is more information to be had than from the main effects by themselves. In other contexts, this would signal a moderation effect of one variable on the other. Either way, it's potentially important information that generally shouldn't be ignored.

Good luck with your work.