Lasso regression coefficient inperpretation

조회 수: 3 (최근 30일)
Tiago Dias
Tiago Dias 2019년 4월 18일
편집: Tiago Dias 2019년 4월 18일
Hi,
I got my data (predictors) and my response variable (Y_MOTOR) and I want to interpret the results when I build a Lasso model.
load('predictors.mat')
load('Y_MOTOR.mat')
[b,Stats] = lasso(predictors,Y_MOTOR,'Lambda', [0.001 0.01 0.1 1 10],'Standardize', true);
The b is a 41x5 matrix, each column represent a value for lamba. Lets say the best model is provided by lambda = 0.001 since is has the lowest value of MSE.
I got b values 0, some very close to zeor (4E-4 or E-3) and then I got 0.1 or 2 or 3 values.
My question is how to rank the regression coefficients in order to say X1 is the more important variable, after X4, etc can i just rank the absolute value of the regressors since they are at the same level because they were calculated with standardized data. Or is there any better metric or methodology?
Thanks for your time

답변 (0개)

카테고리

Help CenterFile Exchange에서 Regression에 대해 자세히 알아보기

Community Treasure Hunt

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

Start Hunting!

Translated by