Degree of Omitted Variable Bias -psacalc

[Priyoma Mustafi]

Hello,

Im Im ing find the the relative extent of the Bias suing psacalc (Emily Oster 2017). I've a household fixed effect (hhid) regression using xtreg.


My regression command is as follows

xtset hhid
xtreg y x1 x2 x3 [pweight=sampleweight], fe cluster(hhid)


This regression command works fine and I get the results on the stata window .
Next I try to do psacalc for independent variable x1 with relevant delta and beta values respectively.

psacalc beta x1, delta(1)

But I get an error "Last test not found" . I get the same error for psacalc delta.

However when I repeat the same thing for a fixed effect regression that is not weighted (ie without pweight) I don't get any such error.

Can someone please suggest how psacalc works for weighted regressions?

Best,
Priyoma

[William Lisowski]

The psacalc command is a user-written extension to Stata. While there may be members here who are familiar with it and can respond, I note that the output of help psacalc gives two email addresses to contact for support. You might be best off addressing yourself directly to them.

I was able to create the following example modifying an example from the help psacalc output; it reproduces your error message - and it works if the weights are omitted. Perhaps the author will find it a useful starting point, or else someone on Statalist familiar with psacalc may be able to use it to offer some suggestions.

webuse nlswork, clear
xtset idcode
generate wt = runiform(.8,1.2)
by idcode: replace wt = wt[1]
xtreg ln_wage grade age c.age#c.age ttl_exp c.ttl_exp#c.ttl_exp ///
    tenure c.tenure#c.tenure 2.race not_smsa south [pw=wt], fe
psacalc beta south

[Anna Jongma]

Hello,

I was wondering if anyone has found a solution for this issue yet?

Kind regards!