What test should I use for my data?

Question

I have one categorical predictor coded as 0 or 1 for a sample of about 290. Based on some research, I want to control for 4 covariates. One is continuous (age) and the rest are categorical (gender, ever having used certain drugs as either yes or no).

My outcome is number of violent acts - this is what I'm having trouble with. According to a professor at my university this isn't a true continuous variable as it is an interval variable. This confused me as I've used interval variables as continuous in other tests.

What type of test would I use if my outcome was continuous? Do I need to use a different one for my interval data?

I am using SPSS to analyze my data. After finding base associations (using t-tests) between my predictor and outcome and covariates and outcome, I tried to go further. I did a univariate analysis of variance (under general linear models) with my predictor as a fixed factor. This seems logical as there's a section for covariates, but I honestly have no idea. Any guidance would be much appreciated.

Answer 1

If you have counts of a type of event as your dependent variable then you should use an analysis that is designed for that type of dependent variable. Poisson regression is the simplest choice to start. Your predictor of main interest and the covariates would be the independent variables, similarly to the standard multiple linear regression you might perform with a truly continuous dependent variable, but they now help predict the probability of an event's occurrence.

Several Cross Validated pages explain the advantages of Poisson regression or related count-specific analyses, for example here and here. Unless you have so many events that your dependent variable is very close to continuous, then a multiple regression treating your dependent variable as having Poisson, overdispersed Poisson, or negative binomial characteristics would seem to make the most sense. (These are examples of generalized linear models; be careful with the choice of terminology.)