stepwise, pr(.05): reg Question: How to keep certain variables from being removed during the selection process

[Joe Korkey]

I am doing the following:

Code:
stepwise, pr(.05): reg Yvariable Xvariable1 Xvariable2 Xvariable3 Xvariable4 Xvariable5 ..... XvariableN

If I think moderators 2 - 6 (Xvariable2 - Xvariable6) are theoretically important and should always remain in the model regardless of statistical significance, is there a way to make those stay in the model throughout the selection process?

Thanks!

[Ken Chui]

Use the command -help stepwise- and then study these two examples on lockterm1:

    Backward selection; force weight to be included in model
        . stepwise, pr(.2) lockterm1: regress mpg weight displ gear turn headroom i.foreign price

    Backward selection; force weight and c.weight#c.weight to be included in model
        . stepwise, pr(.2) lockterm1: regress mpg (c.weight##c.weight) displ gear turn headroom
            i.foreign price

[Carlo Lazzaro]

Joe:
the usual cautionary tale applies

https://www.stata.com/support/faqs/statistics/stepwise-regression-problems