Some help for moderation analysis in SPSS?

Hey everybody!

Im stuck with my psychology masters thesis. Im studying if narcissism and a level of education moderate peoples ability to evaluate their own logical reasoning skills. I assume that narcissism effects the congruence of evaluation and probably makes people to overestimate their skills etc. I also assume that higher education makes people to evaluate their skills more accurate.
Nature of the data - independent and dependent variables

I have data of 256 people. Independent variable is test scores in Raven - a test of logical reasoning. Moderators are narcissism and the level of education (4 levels). Im doing two separate settings: with narcissism as moderator and with education as a moderator. People evaluated their skills in 1 to 5 scale, 5 being the best score. So idea is to predict self-evaluation score with a performance in Raven and to see if narcissism effects the relationship in between them.

What confuses me is that I have a dependent variable that is categorical with five categories. Independent variable (test score) is continuous and moderator is continuous. Can I use linear regression with these variables?

In second setting I have the same categorical dependent variable and continuous independent variable but my moderator is categorical with 4 levels. Is linear regression possible now?

Narcissism = continuous
Level of education = 4 levels (elementary, high-school, school of applied scienses and university)
Raven tests scores = continuous
Self-evaluation of skills = 5 levels

Maybe this has been asked before but Im new here :D

I really would appreciate help here, thank you!


New Member
What do you mean two separate settings?

Can you describe what you mean by self-evaluation of skills? You say that it is categorical, but your description gives a response scale of 1-5. Are these 1-5 evaluations of multiple skills (i.e., separate DVs)

With a continuous DV, linear regression with interaction terms would fit your design.
With a categorical variable, you are looking at ordinal or multinomial logistic regression.