exposure/offset & weights in jwdid method(poisson)

[Ian Wong]

Good morning, everyone! ^_^

When we use jwdid method(poisson), can we do exposure/offset(var) as we do in poisson command? I haven't found it in help files, so just writing to confirm if there's anything equivalent.

Btw, can we use aweight in jwdid? I only find pweight.

Thank you so much for sharing your ideas!

[Ian Wong]

Dear Dr. Rios-Avila FernandoRios, If you are interested, could you please share some ideas? Thank you so much!

[FernandoRios]

no aweights, unless you change code
and look into jwdid options. what you ask is in the helpfile

[Ian Wong]

Thank you very much!
Take dollhill3.dta for instance.
I previously thought m1 m2 and m3 should produce the same results since logA - logB = log(A/B). However, m3 is different! By chance, I found m4 generates the same results as m1 and m2, except for N (# of obs) at the end of the table! Interesting! Seems that poisson command does some transformation when we use offset and exposure.
use D:\stata\dollhill3, clear
gen l_pyears = ln(pyears)
gen pctDeaths = deaths / pyears
eststo m1: poisson deaths smokes, offset(l_pyears)
eststo m2: poisson deaths smokes, exposure(pyears)
eststo m3: poisson pctDeaths smokes
eststo m4: poisson pctDeaths smokes [fw = pyears]
esttab m*

Back to jwdid, I didn't find offset or exposure for method(poisson). As using % as y but omit offset/exposure is not equivalent to using count as y + offset/exposure (shown above, m3 is different) and jwdid only allows for pweight, I guess we may not be able to use jwdid in the case when we have to use offset/exposure (I hope we can or there is some savvy equivalent)...
In the examples in the helpfile, there's one jwdid emp lpop... but lpop here is a covariate and is used for interaction. I guess it's not equivalent to offset/exposure.

Hmm, what I have currently... Hope there are more ideas~~Thank you, everyone! Best regards~

[FernandoRios]

Just add the option at the end of jwdid, as you would do with poisson.

[Ian Wong]

Oh, I just tried - we can further add offset(var) at the end of jwdid..., method(poisson). Then the beta of the var will be constrained to be 1. That's great! Thank you so much!