Docs: Sun Abraham Aggregation with Weights

[s3alfisc]

Hi @ the fixest team,

Over at pyfixest, @shapiromh is working towards adding weights to the Sun-Abraham estimation routine's aggregation method that is modeled afterfixest::aggregate().

fixest::aggregate()'s doc has a use_weights function argument that is explained in the docs as "Logical, default is TRUE. If the estimation was weighted, whether the aggregation should take into account the weights. Basically if the weights reflected frequency it should be TRUE."

So whenever a user is supplying weights for a sunab estimation, fixest defaults to frequency weights in the aggregation step. This is also conformed by taking a look at the code.

But does the feols(sunab()) call with weights not produce a vcov based on a precision weight interpretation? I.e. the number of observations used is N = dim(data)[0] and not N = sum(weights), producing an inconsistency with the aggregation / using a mis-computed vcov?

I have a hunch that @grantmcdermott might have thought about this before, so I am shamelessly tagging you =)

Any help or pointers are much appreciated!

Best, ALex

[grantmcdermott]

Oof, sorry @s3alfisc I actually haven't thought about this. Would require more mental bandwidth than I can allocate rn as I've got family in town and a whole bunch of other deadlines. Please tag me again if you find some resolution in the meantime.

P.S. Do you mean N = dim(data)[1]?

[s3alfisc]

Thanks Grant!

Yes, of course N = dim(data)[1], I am mildly embarrassed 😅

I will think about this a little more and keep you updated here in case I manage to come to any conclusions.