Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Apurba Shee <shee.apurba1@gmail.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | Re: st: xtfrontier tvd |
Date | Sat, 7 May 2011 21:33:40 -0400 |
Hi Scott, Thanks for the idea. I tried rescaling the independent variables (logindepvar - mean(logindepvar)) ... still convergence issue. any idea of other scaling? I am getting very low technical efficiency (around 10%) for the other sectors. Just to double check, does this prediction command (predict te, te) give technical efficiency or technical inefficiency? Apurba On Fri, May 6, 2011 at 4:05 PM, Scott Merryman <scott.merryman@gmail.com> wrote: > You can specify more than one technique so that ML will switch between > the algorithms: > > . webuse xtfrontier1,clear > > . xtfrontier lnwidgets lnworkers lnmachines , tvd difficult tech(nr dfp) > > In the sectors where are you having problems you might try to rescale > the variables so that are all on about the same scale. > > Scott > > > On Fri, May 6, 2011 at 1:29 PM, Apurba Shee <shee.apurba1@gmail.com> wrote: >> Hi Scott, >> I have tried the –difficult- option that does not help. But when I use >> other maximization technique; for bhhh it shows bhhh requires >> observation level scores, and for dfp and bfgs it shows stepping has >> contracted, resetting dfp/bfgs Hessian, discontinuous region >> encountered cannot compute an improvement. I used xtfrontier lndepvar >> lnindepvars, tvd diff technique(bhhh). Do you know how to make these >> techniques work. I appreciate your help. >> Apurba >> > > * > * For searches and help try: > * http://www.stata.com/help.cgi?search > * http://www.stata.com/support/statalist/faq > * http://www.ats.ucla.edu/stat/stata/ > * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/