LASSO regression survey design

AkemiW25 — Wed, 21 Apr 2021 15:34:10 GMT

Hello,

I am trying to use LASSO regression to select micronutrient and lifestyle variables. I learned to apply the glmselect. My challenge is when using the NHANES complex survey data, does the weight statement alone can capture these design parameters or not.

could not find a single example code/ training to do LASSO for complex survey structures in SAS

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7551912/pdf/nutrients-12-02652.pdf (NHANES, LASSO, using R software)

My outcome of interest is a continuous variable, predictors (29- both continuous, and multilevel categorical)

Sub-population sample size is around 4000. I am new to LASSO regression, thus appreciate your advice.

When we use surveyreg, it allows weights, cluster and strata but glmselect only has weight option. If I use the same weight (only) as used in the surveyreg would that be sufficient?

proc surveyreg data= NHANES2001_2002;
weight wtint2yr;
cluster sdmvpsu;
strata sdmvstra;
class female;
model pad630 = female / noint solution vadjust = none;
contrast 'comparing males and females' female 1 -1;
format female fm.;
run;

Re: LASSO regression survey design

gcjfernandez — Fri, 23 Apr 2021 12:53:42 GMT

SAS Survey procs currently do not support any automatic variable selections including LASSO methods.

topic Re: LASSO regression survey design in SAS Data Science

LASSO regression survey design

Re: LASSO regression survey design