Thank you for you reply.

What I want to know is the factor which will affect pregnancy termination. I need the variable v228(ever had a terminated pregnancy).
In IR file, women who have never had a child is also included. I think It will underestimate the results.
In BR file, each women at least has a child, wether the child is alive or not. So I consider to use BR file.
First, I restrict each BR file to women level. i.e., each caseid is unique(each caseid means one women).
Then I used population to de-normalized the weight and pooled all rounds together.
Is it right?

In you reply, you mean, after I pooled the de-normalized data together, I can simply to do "reg Y X" rather than "reg Y X [pw=wgt_new]",right?