Re: [R] all subsets for glm

> If you actually want to find the best subsets, you can get a good > approximation by using leaps on the weighted least squares fit that > is the last iteration of the IWLS algorithm for fitting the glm. > > Running regsubsets witha reasonably large value of nbest and then > refitting the top m

Re: [R] all subsets for glm

If you actually want to find the best subsets, you can get a good approximation by using leaps on the weighted least squares fit that is the last iteration of the IWLS algorithm for fitting the glm. Running regsubsets witha reasonably large value of nbest and then refitting the top models as

Re: [R] all subsets for glm

> Of all the dangerous ways of doing this and getting confusing results, > gl1ce in lasso2 should be the least risky. Thanks Dieter. In case an exhaustive search (all subsets) remains infeasible, I'll include a shrinkage method for sure. Looks like glmpath could be useful here. Best, Harald ___

Re: [R] all subsets for glm

Harald von Waldow wrote: > > For the purpose of model selection I am looking for a way to > exhaustively (and efficiently) search for best subsets of predictor > variables for a logistic regression model. > Of all the dangerous ways of doing this and getting confusing results, gl1ce in lasso2

[R] all subsets for glm

Dear R-users, For the purpose of model selection I am looking for a way to exhaustively (and efficiently) search for best subsets of predictor variables for a logistic regression model. I am looking for something like leaps() but that works with glm. Any feedback highly appreciated. -- Harald