SUGGESTION: Add SELECTION=LASSO capablity to PROC PHREG
A short list of references:
Tishirani, R. (1997) "The Lasso Method for Variable Selection in the Cox Model", Statistics in Medicine, Vol. 16, 385-395 http://statweb.stanford.edu/~tibs/lasso/fulltext.pdf Zhang, Hoa Helen and Lu, Wenbin(2007) "Adaptive Lasso for Cox's proportional hazards model", Biometrika 2007 pp. 1- 13 http://www4.stat.ncsu.edu/~hzhang/paper/alasso.pdf Sohn, Insuk, Kim Jinseog, Jung Sin-Ho, and Park, Changyi(2009) "Gradient lasso for Cox proportional hazards model", Bioinformatics Volume 25, Issue 14, ppg. 1775-1781 http://bioinformatics.oxfordjournals.org/content/25/14/1775.full
+1 for this suggestion!
This had already been implemented in the R package glmnet, although only cross-validation (not AIC, BIC or many of the other criteria available in PROC GLMSELECT for linear regression) can be used to select the optimal value of the tuning parameter.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.