saving p values of regression results using loops?

[Jannic Cutura]

Dear All,
I am using a difference-in-difference setup to evaluate a treatment effect. Unfortunately it is not so clear, when the treatment started. I am working on a qualitative solution by reading through relevant documents, but I also want to work on a quantitative solution. I do know that the treatment started between 18745 and 378 days later. To this end I use the following for loop:

    log on
    forvalues i =1(3)378 {
        di "We are at i = " `i'
        gen treattime = 0
        replace treattime = 1 if date >18745+ `i'
        gen did = treated*treattime
        reg ret did treated treattime, robust
        //reg ret did treated treattime i.bankid, cluster(bankid)
        drop treattime did
        }
    log off

My log reports contain the regression tables and I can see how the p value of the did variable starts to decrease after a while. Now I want to try different specifications and ideally save the p value of the did regressor in a vector where I can relate it to the step i, I am at. Later I would want to export that vector (for example to excel or csv).
Seems like not a standard-Stata-task.

I am aware that this is the kind of data-mining excercise that should not be encouraged, but at this point I am really just trying to understand my data better.

Any advice is appreciated!

Best,

[Maarten Buis]

This Statatip is about how to extract the p-values: http://www.stata-journal.com/article...article=st0137

[Clyde Schechter]

As an aside, don't do

gen did = treated*treattime
reg ret did treated treattime, robust

do this instead:

reg ret i.did##i.treated, robust

In your particular situation, it may not make much difference, but it is best to get in the habit of letting Stata create interaction terms by using factor-variable notation, because most of the time you will want to run -margins- afterward to help interpret the results.