Removing fixed effects for an irregular time period

[Laurence Selwyn]

Hi all, first time posting here but have made use of the stuff you've put up here a lot.

I'm working with a dataset of UK general elections from 1922 to the present, at the district level. I want to remove the fixed effects of the particular election to generate a statistic which I will be using in my work later, but am unsure how to proceed. I am trying to counter the effect of bias against incumbent parties and want to observe more accurately the voting preferences of the constituents preferences. I am not doing any regressions as of yet so I am unsure how to remove these fixed effects.

Any help that could be provided would be invaluable

[Joro Kolev]

At this level of generality (what statistic? from what you want to remove this fixed effect?) it is hard to say anything in particular.

But lets say that you have a variable X that you observe for each district, and for each election. You can generate the deviation of X from the election average of X, and this will "purify" X from the election fixed effect. Something like this

egen Xmean = mean(X), by(election)
gen Xdemeaned = X - Xmean

Now you can make comparison in terms of Xdemeaned and it will be free of election fixed effects.

[Laurence Selwyn]

Yeah, that's the sort of thing I was looking for. I was just being a bit dumb and not thinking it through a lot. Thank you.