Event Study - `i' not changing

[Jacob Grant]

Hi,

I am trying to run the code
"forvalues i = 1(1)319 {
reg return mkt_rtn if id == `i' & estimation_window == 1
predict p if id == `i'
predict std_err_residual, stdr replace SE_residual = std_err_residual if id == `i' //& event window == 1
}"

but I get no observations. I believe that the problem occurs with "id== `i' as I do not see variables changing in the dataset.

[Andrew Musau]

If you have cross-sectional data, then each id is an observation. You cannot run a regression using a single observation. But the error then should be "insufficient observations". Here, you are looping over consecutive ids numbered 1 to 319. If an id is missing, you will get a "no observations" error. Safer to collect the existing ids fulfilling your criterion:

qui levelsof id if estimation_window == 1 & id<=319, local(ids)
foreach i of numlist `ids'{
    reg ...

}

[Clyde Schechter]

To Andrew Musau 's excellent advice I will add:

Inside the loop, you create variables p and std_err_residual, both using the -predict- command. But when you reach the second iteration of your loop, those variables will already exist, having been created in the first iteration. Stata will therefore refuse to carry out the -predict- command and will halt with an error message saying that variable p already exists.

The solution to this is to add -drop p std_err_residual- at the end of the loop.