04-24-2014 06:43 PM
okay... the equation of my study is
HamiltonDepressionScore=week lndesipramine lnimipramine endogenous week*lndesipramine
i want to see if gender(categotical) will be confounding in the relationship between HamiltonDepressionScore and lndesipramine for instance. i want to know if theres something i can observe to test this, like the pvalue or the standard error of something. i dont know if that explains anything
04-25-2014 10:44 AM
Maybe I am not understanding well, but I think you are looking to see if there is an interaction between gender and the other variables. There are several PROCs that will enable you to do this, so my next question would be whether you can post your current code (just the analysis procedure part would do).
From that, we can make suggestions to get gender into the model and see if the effects of the other variables are consistent across gender. It would also help to know a bit more about the dependent variable--I suspect that it is an ordinal variable, so the best analysis probably ought to consider that.
04-24-2014 10:56 AM
There are a host of different models that are available in SAS that are applicable to issues involved in the analysis of longitudinal data. And, as Steve Denham points out, the term "confounding" has several meanings in a statistical sense. I further suspect that given the nature of your data and the issues you are investigating that the applicable theory driving the research also has something to say about the "confounding" and how it should be addressed in a statistical sense.
04-29-2014 07:45 PM
In addition to not referencing gender anywhere in the existing equation, I also note that there does not appear to be any method of controlling for baseline status on the Hamilton Depression Score (assuming that some change-related modeling is central to the analysis). That could very well be a function of the proc that is being used, however.