LASSO, Ridge & Elastic Net...

[miked]

Please can I ask two small questions regarding the use of LASSO, Ridge and Elastic Net options that you've added recently?

Firstly, is it possible to restrict the sum of the coefficients to equal some value (say, 1)? (If I try the approach usually used in OLS - e.g. setting y=c(1)+c(2)*x1+(1-c(2))*x2), then I find it crashes EVIEWS.

Secondly, is it possible to restrict the sign of each of the coefficients (say, to be >0)?

Thanks for your help.

Regards
Mike

[EViews Rebecca]

At the moment it's not possible to restrict the sign on the coefficients. Are you able to post the workfile that's crashing?

[startz]

Can you go

Code: Select all

y = exp(c(1)) + exp(c(2))*x

?

[Yangka]

Dear Miked and Rebecca, I could not create new post.
So I am using your post threads seeking your support on ridge, lasso and elastic net - I am new to these regressions.

1) is regressor transformation (that comes up in the options tab) always necessary?
2) with transformed regressors, how do we interpret the coefficients that appear in the result?
3) how do we check if the results are reasonable? Is it enough just to check that the training data set lies below the test data set?
4) how do we confirm if using these regressions have fixed the multicollinearity issue faced in OLS?

I have 8 variables each with 30 yearly data.

Look forward to hear from you.
thanks for your time