减少变量逻辑回归-MATLAB

Question

I have a very large matrix with 37 subjects and 1770 variables that has two possible outcomes (0/1). I am trying to perform logistic regression, but naturally the model is overparameterized. I have tried forward selection with sequentialfs, and I have also tried Lasso (even though this is not a good approach, since the data in each column is not normally distributed).

Does anyone have a good idea to what methods, I can use in MATLAB to remove the redundant variables and complete my logistic regression?

Cheers!

Answer 1

您可以使用SVD或PCA通过删除相关变量来减少参数数量。