How to compare differences between groups in predictors of an outcome variable?

#1
Hi everyone,

I have one outcome variable (A), many predictor variables (X,Y,Z,..) and three groups (clinical groups I, clinical group II, and one control group). I would like to see if there are differences in predictors of A between groups (at least between the two clinical groups).

I was thinking fist to do a standard regression separately for each group (if I have big enough sample for each group), then probably to use the predictors that are in common and run another regression analyses for the whole group. And then use group as a dummy variable or use a moderation analysis with multiple regressions with interactions? But I am not sure if this will give me what I am looking for. Does anyone have any idea if and how can I do that?

Many Thanks,
Valia
 
Last edited:

hlsmith

Not a robit
#2
I would start by just making a table with groups as the columns and rows as the rows. I would then fill in the table with descriptive statistics. That may be good enough. since you will be reporting the differences between the groups in a table of the model results. Some times people run too many tests that aren't necessary, especially if you are already controlling for those covariates in the model.
 
#3
Thanks for your reply hlsmith. That's right. I initially plan to see group differences using AVOVAS and correlations. I will then enter into the model only the ones that are significantly correlated. I am not sure though, what will be the next step for comparing the differences in regressions between groups.
 

hlsmith

Not a robit
#4
You have to understand the relationship between the covariates, you can't just use the results of the table since the variables may be multicollinear or interact with each other. You have to use your content expertise when selecting variable for the model.