Asymptomatic/Finite Sample analysis

[Jared Greathouse]

This may not be the forum for this- perhaps Cross- Validated, but I don't have a Stata question as much as I have a stats question.

Whenever I read fancy papers in Journal of the American Statistical Association or Journal of Econometrics or similar outlets, I frequently hear of things like "finite sample analysis" or "asymptomatic analysis". And I know what they are, they're essentially tests or processes we go about to uncover qualitative info about quantitative estimators.

But even so, I've never done this. They CERTAINLY have never taught us this in my graduate courses in political science, and I'll bet my next three paychecks that my current Ph.D program in public policy won't have such classes.

I wanna be an applied econometrician, but I also wanna have a strong mathematical background on what I do too. Does anyone know of any resources, textbooks perhaps, which go over asymptomatic analysis/finite sample analysis? I'd like to learn more about it, if possible.

[Leonardo Guizzetti]

I've never done this either and I wasn't taught about it in my education, so I may be off base a bit here. Usually I see one of two things when reading biomedical and epidemiology literature. The first is to use permutation tests when asymptotic results can't apply, which may not truly be finite sample methods. Or else adjustments are made to variance estimates to account for finite sample estimation. In this vein, you may want to look into survey analysis methodology for finite sample analysis.

[Joao Santos Silva]

Dear Jared Greathouse

I suggest that you try to audit a course in econometrics, but this is also a good point to start from:

https://www.ssc.wisc.edu/~bhansen/econometrics/

Best wishes,

Joao

[Jared Greathouse]

Joao Santos Silva I've taken... I think three courses in GLMs, two on causal inference (my specialty), but I've never taken a normal econometrics course in an econ department. So, perhaps I'll take the doctoral Advanced Econometrics 2 course at my school, it seems like it'll be lots of fun and allow me to test the new causal inference estimator I've written.