help > Multiple Linear Regression
Showing 1-2 of 2 posts
Display:
Results per page:
May 23, 2024  04:05 PM | Natalie Bell
Multiple Linear Regression

Hello,


I am trying to determine how to set up a follow-up analysis. I have significant findings between two groups and I want to see if these findings are dose dependent by adding in a quantative metric. I want to see specifically if there is a positive or negative correlation between one of the groups and this dose metric but the control group does not have this dose metric so I am wondering if this will be possible. How would I go about designing a linear regression model in the NBS accordingly? Would I keep only one group in the design matrix? 

May 25, 2024  12:05 AM | Andrew Zalesky
RE: Multiple Linear Regression

Hi Nicole, 


Given that the control group does not have the dose metric, you may want to consider a follow-up analysis  that only includes the active/patient group. The design matrix would include a columns of ones and column for your dose metric as well as any other covariates. The contrast would be [0 1] (postive correlation) or [0 -1] (negative correlation)


Another approach would be to include both groups and simply use a value of zero for the dose variable of the control group. But this might not make sense, depending on your context. 


Originally posted by Natalie Bell:



Hello,


I am trying to determine how to set up a follow-up analysis. I have significant findings between two groups and I want to see if these findings are dose dependent by adding in a quantative metric. I want to see specifically if there is a positive or negative correlation between one of the groups and this dose metric but the control group does not have this dose metric so I am wondering if this will be possible. How would I go about designing a linear regression model in the NBS accordingly? Would I keep only one group in the design matrix?