Use of "stepwise" or "swboot" commands with complex survey data

[Kavi Singh]

I am performing a linear regression on a dataset from a complex survey design, using backward stepwise estimation. I would also like to explore bootstrap stepwise regression. Do the "stepwise" or "swboot" commands allow for 'svy' specification?

[Rich Goldstein]

when you say "linear regression", I assume here that you mean the -regress- command - if yes, that command allows the use of pweights so you should be fine with stepwise (ingoring the general problems with stepwise)

I don't know anything about, and can't find, the "swboot" command - where did you find it?

note that one of the "vce()" options for -regress- is "bootstrap" - will that suffice?

[Kavi Singh]

Thanks for the reply. The swboot command performs stepwise bootstrap regression (Joanne M. Garrett, 2000. SWBOOT: Stata module to bootstrap stepwise linear or logistic regression models," Statistical Software Components S415601, Boston College Department of Economics). The regress command does allow for 'svy' specification. However, I have not been able to successfully incorporate svy into the "stepwise" command. It appears that "stepwise" does not support svy.

[Rich Goldstein]

which is why I suggested using pweights