[Badar Khalid]

Thank you Clyde for your reply and clarification.. Yes I have removed the comma and it worked fine now.

In fact, I am looking to get P-values corrected for country level clustering, am I getting it right when I add vce (cluster CountryID) at the end of the command or I need to add some other code at the end to get the P-values corrected for country level clustering?. Thank you

[Clyde Schechter]

I assume here you are referring to the -margins- command you asked about in #29. The regression you ran before -margins- already specifies the vce(cluster CountryID) option. That is where the country-level clustering gets taken into account. The -vce()- option of the -margins- command does something different, that has nothing to do with clustering within the data, and it will not accept vce(cluster CountryID). For your purpose, you do not need to specify any vce() option on the -margins- command.

[Badar Khalid]

Thank you Clyde for your reply and for your clarification. I am sorry I have not explain my recent query well. I mean her if I run the following command, Do I need to add some other code to get the P-values corrected for country level clustering?. Thank you.

xtprobit IsBoardBusy  l_IO_Domestic l_LogTotalAsset l_SalesGrowthRate, vce (cluster CountryID)
margins, dydx (l_IO_Total l_LogTotalAsset l_SalesGrowthRate) atmeans

[Clyde Schechter]

No, you don't have to add anything else. The vce(cluster CountryID) in the -xtprobit- command takes care of it.

There is also a -vce()- options for -margins-, but it has nothing to do with clustering, and you don't need to, and should not, use it.

[Badar Khalid]

Thank you Clyde for your valuable comment.

I have also one query with regard to regressing whether the changes in explanatory variables drive changes in dependent variable. In my case, I would like to run pooled OLS regression to investigate whether the changes in l_IO_Domestic (t-2) - (t-1) drive changes in IsBoardBusy(t)-(t-1).

Can you please advise me with the command as I have 7 years data from 2006-2012. Thank you

[Clyde Schechter]

You clearly are doing a lot of analyses like this, so it behooves you to become familiar with the time series operators that Stata makes available. Start with -help tsvarlist-. Then go to the user manual section on this topic. (At the top of the help for tsvarlist there will be a clickable link to the corresponding manual section.) These operators work nicely with the various regression commands in Stata and can be applied to both outcome and predictor variables (although, as we saw, there are snags here and there.) Try working this one out on your own first. Between the help and the manual section on time series operators and using what has come earlier in this thread as a model for what can be done with them, I think you can figure it out. So give it a try, and if you run into difficulties, post back with what you've tried and how Stata responded (and some relevant sample data).

[Badar Khalid]

Thank you Clyde.. I will go through what you have recommended me to do. I really appreciated all your previous valuable comments which helped me a lot.