MATLAB Answers


Lasso regression coefficients interpretation

Asked by Tiago Dias on 18 Apr 2019
I want to built a lasso model for my data (predictors) to predict Y_MOTOR.
[b,Stats] = lasso(predictors,Y_MOTOR,'Lambda', [0.001 0.01 0.1 1 10],'Standardize', true); % standardize the data
So b is 41x5 dimension, each column represent a value for Lambda. Some b values are 0 others are non-zero. Lets say the best model is for Lambda = 0.001, since it gives the lowest MSE. How can I rank my regression coefficients in terms of importance? How can I say, form the non-zero which are important and which are non-important, since i got some that are 4E-3 or E-4 compared to those who are 3 ou 4. Is there any metric? or should i stich with them even if they are very close to zero.
Thanks for your time.


Sign in to comment.

0 Answers