Why -xtoverid- does not seem to support -fvvarlist-?

[Carlo Lazzaro]

Dear All,
I have just come across what follows, related to the user-written command -xtoverid- (net describe xtoverid, from(http://fmwww.bc.edu/RePEc/bocode/x).

. use "http://www.stata-press.com/data/r14/nlswork.dta", clear
(National Longitudinal Survey.  Young Women 14-26 years of age in 1968)

. quietly xtreg ln_wage i.race##i.grade c.age, re vce(robust)

. xtoverid
1b:  operator invalid
r(198);

. quietly xtreg ln_wage i.race i.grade c.age, re vce(robust)

. xtoverid
1b:  operator invalid
r(198);

. quietly xtreg ln_wage race grade c.age, re vce(robust)

. xtoverid

Test of overidentifying restrictions: fixed vs random effects
Cross-section time-series model: xtreg re  robust cluster(idcode)
Sargan-Hansen statistic  54.070  Chi-sq(1)    P-value = 0.0000

Is there any way to workaround this drawback? Thanks.

[Ariel Karlinsky]

Been discussed several times:
http://www.stata.com/statalist/archi.../msg00707.html

Also note that xtoverid doesn't support modern stata "labels" for variables such as o. (omitted) & b. (base)
The solution to this is to estimate using xi prefix, than estimating again without the xi prefix and excluding from the estimation command any dropped variables.
I'm pretty sure there's a "nice" way of doing this by getting the e(b) matrix, dropping all elements with o. b. prefix (or all elements evaluated to 0?) and then re-estimating.
I must say that I tried looking into a way of doing this, but my stata matrix manipulation skills are very lacking.

[Carlo Lazzaro]

Ariel.
thanks a lot for pointing me out to a thread i wasn't aware of.

[Ariel Karlinsky]

Maybe we can use this tread to encourage Mark Schaffer to update xtoverid to use factor variables

[Carlo Lazzaro]

Ariel:
great idea!