Interaction term variable as a control variable.

[geeta dangwal]

Dear Members, I am using stata13 and panel data to study role of financial development in economic growth.
I have one question regarding interaction term in fixed effects method.
I have introduced two interaction term by multiplying private credit with banking crisis(dummy variable) and other by multiplying private credit with region dummy.
1.) Is it accurate to consider banking crisis dummy and region dummy as control variables as well?

2.) When i used banking crisis dummy as control results were almost same. but region dummy has been omitted so again result remain unchanged. what could be possible explanation for omission of region dummy?
Best Regards.

[Clyde Schechter]

First, since you are using a modern version of Stata, you would be better off using factor variable notation (-help fvvarlist-) instead of multiplying variables to create your own interaction terms.

Second, the general rule is that you should include each of the components of the factor variable as effects in the model as well.

Third, in panel data, if one of these components is constant within panel (which, I suspect will be the case for region), then it is collinear with the panel fixed effects and it is automatically dropped, just as Stata would resolve any colinearity by dropping one of the variables.

[geeta dangwal]

@Clyde Schechter...Thank you very much for your comment.

[Carlo Lazzaro]

Geeta;
just to top-off Clyde's superb advice, why don't you post what you typed and what Stata gave you back (as per FAQ), so that all the listers can see the issue and reply (more) positively? Thanks.

[Richard Williams]

I agree with the excellent advice given by Clyde and Carlo. I'll just add that sometimes in an FE model you do NOT want to include all the main effects. This is wildly counter to what my usual practice is but see Jeff Pitblado's posts in this thread:

http://www.statalist.org/forums/foru...fe-and-margins

[geeta dangwal]

@Carlo Lazzaro....could you please tell me how to post results here? shall i attach pdf of stata ouput or logfile? Actually i read FAQ but could not find guideline regarding posting results.

[geeta dangwal]

@Richard Williams.....I regressed each interaction term in different regression equation to get their effects. Is not it right approach?

[Clyde Schechter]

Don't attach anything. First set up a code block: Open the advanced editor. Click on the underlined A button at the top of the editor Then click on the # button. Code block delimiters will appear. Now go to your log file (or Stata's Results window), highlight the part of the results you want to show, and copy it (however your operating system does copy). Then click in between the code delimiters back here in the Forum editor, and paste (however your operating system does that) there.

You can type other material above or below the code block by way of explanation. Then post your reply.

[geeta dangwal]

Clyde Schechter, sorry to bother you again. Actually I am using Stata/IC 13.1 versions. After clicking advance editior, option of "convert line ending to mac and windows format" came.

[geeta dangwal]

@Clyde Schechter....sorry I misunderstood your post. I was searching for advance editor in Stata :-)