Dynamic heterogeneous panel using the xtprmg command - Hausman test

[Jala Youssef]

Hi everyone,

I have a panel dataset of countries and years. I am undertaking a dynamic heterogeneous panel using the xtprmg command (pooled mean group pmg, mean group mg and dynamic fixed effects dfe). I have a list of explanatory variables of interest and I would like to run the command below:

local ivar1 interest1
local ivar2 interest2
local ivar3 interest3
local ivar4 interest4
...
local ivarn interest

forv i=1/n {

xtpmg d(y x1 x2 `ivar`i'' ), lr(l.y x1 x2 `ivar`i'' )
estimates store pmg
drop __ec
xtpmg d(y x1 x2 `ivar`i'' ), lr(l.y x1 x2 `ivar`i'' ) mg
estimates store mg
hausman mg pmg, sigmamore
xtpmg d(y x1 x2 `ivar`i'' ), lr(l.y x1 x2 `ivar`i'' ) dfe
estimates store dfe
hausman mg dfe, sigmamore
outreg2 [pmg mg dfe] using "$temp_dir\results.xls"
}

I would be grateful if you could advise on how to report the p-values of the two Hausman tests in the outreg2.

Many thanks in advance.

Jala

[Andrew Musau]

webuse nlswork4, clear
xtreg ln_wage age msp ttl_exp, fe
estimates store fixed
xtreg ln_wage age msp ttl_exp, re
estimates store random
hausman fixed random, sigmamore
local hausman= cond(`r(p)'<0.01, string(`r(chi2)', "%9.3f") + "***", ///
                   cond(`r(p)'<0.05, string(`r(chi2)', "%9.3f")+"**", ///
                       cond(`r(p)'<0.1, string(`r(chi2)', "%9.3f")+"*", ///
                           string(`r(chi2)', "%9.3f"))))
outreg2 [fixed] using myfile, replace addtext(Hausman stat., "`hausman'")
outreg2 [random] using myfile, append

Res.:

[Jala Youssef]

Thanks a lot Andrew Musau for your prompt and useful response.

Sorry to bother you again, I have another clarification question: I prefer to report the p-value of the test rather than its statistic. Could you please advise on the relevant code?

Many thanks once again.

Jala

[Andrew Musau]

webuse nlswork4, clear
xtreg ln_wage age msp ttl_exp, fe
estimates store fixed
xtreg ln_wage age msp ttl_exp, re
estimates store random
hausman fixed random, sigmamore
local hausman= cond(`r(p)'<0.01, "p<0.01", "p=" + string(`r(p)', "%4.3f"))
outreg2 [fixed] using myfile, replace addtext(Hausman test, "`hausman'")
outreg2 [random] using myfile, append

[Jala Youssef]

Thanks a lot Andrew Musau . This is super helpful and it works perfectly.

On a separate yet related note, I have a problem with the code I mentioned earlier (please find it below after updating it with the p value of Hausman test). The output file "results.xls" would only show the results with the last ivarn and disregards the results for the rest of the variables of interest (ivar1, ivar2, etc.). I would be grateful if you could advise on how to fix this.

local ivar1 interest1
local ivar2 interest2
local ivar3 interest3
local ivar4 interest4
...
local ivarn interest

forv i=1/n {

xtpmg d(y x1 x2 `ivar`i'' ), lr(l.y x1 x2 `ivar`i'' )
estimates store pmg
drop __ec
xtpmg d(y x1 x2 `ivar`i'' ), lr(l.y x1 x2 `ivar`i'' ) mg
estimates store mg
hausman mg pmg, sigmamore
local hausman= r(p)
xtpmg d(y x1 x2 `ivar`i'' ), lr(l.y x1 x2 `ivar`i'' ) dfe
estimates store dfe
outreg2 [pmg mg dfe] using "$temp_dir\results.xls", replace addtext(Hausman test, "`hausman'")
}

[Andrew Musau]

You are overwriting the estimates while looping. You need:

forval i=1/n {
    xtpmg d(y x1 x2 `ivar`i'' ), lr(l.y x1 x2 `ivar`i'' )
    estimates store pmg`i'
    drop __ec
    xtpmg d(y x1 x2 `ivar`i'' ), lr(l.y x1 x2 `ivar`i'' ) mg
    estimates store mg`i'
    hausman mg`i' pmg`i', sigmamore
    local hausman= r(p)
    xtpmg d(y x1 x2 `ivar`i'' ), lr(l.y x1 x2 `ivar`i'' ) dfe
    estimates store dfe`i'
    outreg2 [pmg1 mg1 dfe1 pmg2 mg2 dfe2 ...] using "$temp_dir\results.xls", replace addtext(Hausman test, "`hausman'")
}

[Jala Youssef]

Thanks a lot Andrew Musau. I really appreciate all your help.

Just a clarification question, in this case just to make sure the "local hausman" would pick the p-value for each ivar regression, do you think I also need to add an 'i' next to it like this: local hausman`i'= r(p) ? if this is the case, should I also reflect in the replace add text in the outreg?

Thank you again so much and apologies for the too many questions.

Jala

[Andrew Musau]

Yes, such that you have:

outreg2 [pmg1] using myfile, replace addtext(Hausman test, "`hausman1'")
outreg2 [mg1 dfe1] using myfile, append

outreg2 [pmg2] using myfile, append addtext(Hausman test, "`hausman2'")
outreg2 [mg2 dfe2] using myfile, append

and so on.