Shuxia Yu wrote:
> Dear all,
>
> I hope to do variable clustering using ``varclus'' in library Hmisc
> to reduce the variables in my data set. And then do linear regression
> on the reduced data set using the clusters obtained in previous step.
>
> However, I don't know how to represent the variables ensembled
> in the clusters. Would you like to give me some suggesitons?
Good possibilities are principal components (e.g. simple pc1 function in
Hmisc) and nonlinear principal components (transcan in Hmisc). Details
are in my book Regression Modeling Strategies.
>
> BTW, it may not a good way to do linear regression on a data set with a
> larger variables. Any hints?
No, it's a pretty good way. -FH
>
> Thank you very much for your consideration on this matter.
>
> Best wishes,
>
> Jinsong
--
Frank E Harrell Jr Professor and Chair School of Medicine
Department of Biostatistics Vanderbilt University
|