Dev,
I was able to replicate this error. When looking at the log, the problem appears to be the call to -fron_p.Calc-
---------------------------------------------------------------------------------------
log: C:\Documents and Settings\scott.merryman\Desktop\foo.smcl
log type: text
opened on: 4 Aug 2004, 06:57:42
. predict te,te
------------------------------------------------------- begin predict ---
<snip>
---------------------------------------------------- begin fron_p.Calc ---
<snip>
- else if "`e(het)'" == "uv" {
= else if "uv" == "uv" {
- tempvar xb_v sigma_v sigma_u
- qui _predict double `xb_u' if `cond', xb eq(lnsig2u)
= qui _predict double if __000000, xb eq(lnsig2u)
too many variables specified
qui gen double `sigma_u' = exp(0.5*`xb_u')
qui _predict double `xb_v' if `cond', xb eq(lnsig2v)
qui gen double `sigma_v' = exp(0.5*`xb_v')
}
}
What believe is the problem is that xb_u temporary variable is not declared before the -_predict double `xb_u' ...-
I have added xb_u to the tempvar line in -fron_p- and now it seems to work fine:
. clear
. use http://www.stata-press.com/data/r8/greene9.dta
. encode state, gen(state_id)
. qui frontier lnv lnk lnl , vhet(capital) uhet(state_id )
. predict te,te
Of course, it is unwise to alter offical Stata programs; hopefully there will be an offical update to this soon.
Hope this helps,
Scott
----- Original Message -----
From: Dev Vencappa <[email protected]>
Date: Wednesday, August 4, 2004 6:28 am
Subject: st: predicting technical efficiency after frontier command: error message
> >>
> >>
> Stata users,
>
> I am running the following command.
>
>
> reg lnrev lntoen lncap lntech lndebt
> mat X=e(b)
> local a=el(X,1,1)
> local b=el(X,1,2)
> local c=el(X,1,3)
> local d=el(X,1,4)
> frontier lnrev lntoen lncap lntech lndebt, diff uhet(lnass dumfra-
> dumother,noconstant) vhet(lnass dumfra-dumother,noconstant) /*
> */from(lntoen=`a' lncap=`b' lntech=`c' lndebt=`d')
> predict CDrev1,te
>
> I am getting my parameter estimates and everything else fine,
> except for the predict command which gives the following error
> message.
> " too many variables specified
> r(103); "
>
> I am basically modelling heteroscedasticity in both the one-sided
> and two-sided error component. In contrast, when I model
> heteroscedasticity in either one of them, then the predict command
> works fine. I have read the manual and it does not say anything
> about possible problems with predict, te when heteroscedasticity
> is modelled for both error components.
> Any explanation why this is happening?
>
> Many thanks
> Dev
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
