Importance of the variable selected by adaptive LASSO

[DavyJones]

Hi,

I have a panel data, for which I try to use the adaptive LASSO model to identify the most important covariates. I am new in this area and appreciate your help. I run the following program (Instead of four covariates mentioned below, I have 45 independent variables).

proc glmselect data=infile;
model y=x1 x2 x3 x4/selection=LASSO(adaptive stop=none choose=bic);
run;

following is the LASSO selection summary in the results window. I wonder what the number effects column is in the following table. Is there any relation between this ranking and the importance of the variables? (I noticed that this ranking is not based on the size of the reduction in BIC)

The GLMSELECT Procedure
LASSO Selection Summary
Step	Effect	Effect	Number	BIC
Entered	Removed	Effects In
0	Intercept		1	474393.91
1	x3		2	474313.58
2	x1		3	474273.03
3	x2		4	473892.46*

Thanks.

[Ksharp · Accepted solution]

I think should check the value  of  BIC-lag(BIC)