(PS I had originally planned on using multiple regression, but now am not sure that's the right method)

- Thread starter bcrider
- Start date
- Tags analisys linear regression methodology

(PS I had originally planned on using multiple regression, but now am not sure that's the right method)

I don't see how you can use categories that are not mutually exclusive for a given unit for a predictor. Or what the value of using such categories would be anyway. It is not clear to me if you running multiple DV at the same time (in which case ANOVA certainly won't work - you need something like MANOVA) or if you are going to run multiple ANOVA's each with a specific DV. In the latter case you test linearity as always, but as always comparing across DV is not simple (or anyhow I don't know a simple way to do this, one might exist).

I think you need to provide more detail on what specifically you are trying to do, the measurement of the variables, and why you are doing what you are.

I am wanting to look at how medications effect change in volume measurements. I'm going to be doing multiple analyses (through a for-loop in R, changing the DV till I get through each volume measurement). The categories are not mutually exclusive in that if I want to compare med class A's effects on volume 1 change to class B's effects on volume 1 change, there are some people not on meds in either class, some people on one, and some people on both. Let me know if you need more details! Thanks!

I may not understand what you are doing, but normally if you want to to compare the impact of treatment on something [here volume] individuals who are not in one of the treatment groups or are in more than one treatment group would not be usable [they would be excluded from the analysis or alternately treated as a control]. Obviously you can not determine the impact of individuals not in one of the predictors while you can not tell the impact of A or B if someone is in both A and B. You might create a third category called, both A and B and determine its impact.

Individuals who recieve no treatement might be considered a control group and counted this way [here you have one variable with 4 levels, in A, in B, in A and B, not in A or B commonly called a control group if you are testing A and/or B].