aweights in binary logit regressions

[George Kariuki]

I have learnt that since Stata 10.1, the use of analytical weights were removed due to their interpretational difficulties. When running a regression while still hoping to correct for sampling issues, what weights should one use? aweights return an error message with Stata 13.

. logit povertydmy hhage hhage2 hhsize educyrs i.gender i.marstat i.resplace [aw=weight]
aweight not allowed
r(101);

[Carlo Lazzaro]

George:
perhaps -pweight- is what you're interested in.

[George Kariuki]

Technically, what do person weights correct for? I see from help pweights that they are the inverse of probability that an observation is there of sampling design..

[Carlo Lazzaro]

George:
from -help weight-:

pweights, or sampling weights, are weights that denote the inverse of the probability that the observation is included because of the sampling design.

.

I guessed it was interesting for you as you stated you were experiencing sampling issue.
Otherwise, take a look at -fweights-, if you have grouped observations.

[Friedrich Huebler]

See this discussion on the DHS user forum for information about weights. The DHS experts recommend pweights.

[George Kariuki]

Really helpful direction. Thank you very much.