rreg

[Tracy Lam]

Hi,
I am attempting to run robust regression in Stata with a small sample size (n=27). I tried looking in the user's manual but don't see anything about whether that is a problem. Does anyone have information on this? Is it OK to run rreg with a small number of observations?

[Nick Cox]

A sample size of 27 is just that and I don't think rreg raises any special issue in that respect.

Apart from that the bigger deal is why you don't use a more modern flavour of robust regression.

See e.g. http://www.statalist.org/forums/foru...reg-vce-robust

http://www.statalist.org/forums/foru...thods-in-stata

[Joao Santos Silva]

Dear Tracy,

Following up on Nick's comment, I would go even further: why do you want to use "robust" regression? If you are worried about outliers, you may be better off using quantile (median) regression.

All the best,

Joao

[Nick Cox]

Joao's point about quantile regression echoes comments in the second thread I cited.