How to reduce overparameterized model for stepwise glmfit?

1 次查看（过去 30 天）
Alayt Abraham Issak
编辑： Alayt Abraham Issak 2019-4-15
I am trying to use the stepwise regression generalized linear model for a dataset with 85 columns and 2055 rows. Many of these colums are booleans (0,1) and some include a good amount of Nan values, which I know is quite a weakness to my dataset :(
I am using the code below to generate my model: (A_train = 85 columns) & B_train (1 or 0)
Model = stepwiseglm(A_train,B_train,'constant','Upper','Linear','Distribution','Binomial','PEnter',0.05,'PRemove',0.10);
However, my result produces no model as all my factors are taken out. I understand that matlab is avoiding to overfit my factors hence, I was plannig on removing some of these factors myself before runnig the stepwiseglm function.
What methods can I use to do so before running my stepwise function? I was thinking of running the correlation betwwen each factor and the predictor (B_train) and then eliminating the negative correlating columns however that may skew my data.
Any sugesstions are welcome! Thank you

请先登录，再回答此问题。

回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Linear and Nonlinear Regression 的更多信息

标签

Community Treasure Hunt

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

Start Hunting!

Translated by