Extract final model from stepwiseglm?

2 ビュー (過去 30 日間)
Jim 2017 年 2 月 9 日
I'm using stepwiseglm to choose predictors from a large predictor suite (as expected/intended). I run the code;
X = [x1 x2 x3 x4 x5 x6 x7 x8]; mdl = stepwiseglm(X,y,'linear','upper','linear', ... 'Distribution','binomial','Criterion','sse');
and the predictors x1, x4, and x7 are selected by stepwiseglm. At this point though, "mdl" is a model with 8 predictors, not 3. So if I try to use mdl in another program that applies the model to a new set of predictors x1, x4, x7, an error occurs that the model "mdl" expects 8 predictors. That is,
X_new=[x1_new,x4_new,x7_new]; y_predicted=predict(mdl,X_new);
doesn't work. My work around is to manually handle it with:
X_selected = [x1 x4 x7]; mdl2=fitglm(X_selected,y,'distribution','binomial'); y_predicted=predict(mdl2,X_new);
My question is whether I can extract the 3-predictor model from mdl directly as a Class GeneralizedLinearModel so that I don't need to manually enter the selected predictors? There is a lot of information available about mdl, but I don't see anything obvious about where to find the "selected model" as a standalone GeneralizedLinearModel class.
Thanks in advance, Jim

回答 (0 件)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by