IV-Regression with ivreg2- Does not save first stage

[Patrick Johannes]

Hello there,

I am instrumenting my governance variable with politics variables. However my code does not save the first stage results.

ivreg2 risk4_w (governance_pillar_score = president congress govenor) company_size_w earnings_smoothing_2_w leverage_w dividend_ratio_w research_development_ratio_w sales_growth_w tangability_ratio_w turnover_rate_w i.sic_2, partial (company_size_w earnings_smoothing_2_w leverage_w dividend_ratio_w research_development_ratio_w sales_growth_w tangability_ratio_w turnover_rate_w i.sic_2) endog(governance_pillar_score) cluster(company_code) first savefirst sfirst
est store _ivreg2_risk4_w
est restore _ivreg2_governance_pillar_score
outreg2 using "IV-Regression", replace word tstat cttop(first)
est restore _ivreg2_risk4_w
outreg2 using "IV-Regression", cttop(second) word tstat

The problem here is the line with est restore _ivreg2_governance_pillar_score

. est restore _ivreg2_governance_pillar_score
estimation result _ivreg2_governance_pillar_score not found

Can somebody help me with that?

[Joro Kolev]

Neither the user written -ivreg2- nor -ivregress- store the first stage.

You need to run the first stage manually, and save what you need this way.

[Andrew Musau]

outreg2 and ivreg2 are from SSC (FAQ Advice #12). Here is an example that saves the first stage.

sysuse auto, clear
ivreg2 price i.rep78 (foreign = weight turn trunk), first savefirst
est restore _ivreg2_foreign
outreg2 using "myfile", replace
ivreg2 price i.rep78 (foreign = weight turn trunk)
outreg2 using "myfile", append

[Pradeep Fonseka]

I got an error for the following program. The error is "invalid syntax r(198);" which seems to be in the last step. I highly appreciate anyone who can help solve this problem. I think Andrew Musau has expertise. Thank you!

* Load example dataset

sysuse auto, clear

* Run IV regression and save first-stage results
ivreg2 price i.rep78 (foreign = weight turn trunk), first savefirst savefprefix(fs)

* Save second-stage regression
est store second

* Restore and store first-stage regression
est restore fsforeign
est store first

* Export first-stage regression
outreg2 [first] using "myfile.rtf", replace dec(2) ///
title("First Stage Regression")

* --- Extract statistics for addstat() ---
scalar cdf1 = e(cdf) // Cragg-Donald F-stat
scalar sstat1 = e(sstat) // Stock-Yogo S statistic
scalar r2 = e(r2) // R-squared
scalar N = e(N) // Sample size

* Manually compute first-stage F-statistic
qui reg foreign weight turn trunk i.rep78 if e(sample)
scalar F1 = e(F)

* Append second-stage regression with full diagnostics
outreg2 [second] using "myfile.rtf", append dec(2) ///
addstat("CD Wald F", cdf1, "SW S stat.", sstat1, ///
"R-squared", r2, "First Stage F-Stat", F1, "N", N) ///
title("Second Stage Regression")

[Andrew Musau]

If interested in #4, see https://www.statalist.org/forums/for...=1748924854453.