st: RE: Frontier code- r(1400) error

[Nick Cox <n.j.cox@durham.ac.uk>]

The main advice on this kind of information is to try a much simpler model. How many parameters are you trying to estimate? How many data values do you have? Is that reasonable? 

Nick 
n.j.cox@durham.ac.uk 

Hande 

I have a problem with the "frontier" code. When I run the code, " r(1400)
initial values not feasible" error occured. I searched the statlist
archive and found the following post:
http://www.stata.com/statalist/archive/2003-05/msg00650.html.
When I apply these steps, I got the r(1400) error again.


Another problem that I faced with is about the distribution assumption. I
want to find the profit efficiency scores, and most of the papers advice
to use half normal distribution but when I use the half normal
distribution, r(430) error occured.( cannot "compute an improvement --
discontinuous region encountered). Then I tried exponential distribution,
and for some years I got the estimation results. Which distribution should
I use for the profit efficiency? Can exponential distribution be used for
the profit efficiency scores?


Here are the codes:

regress lnpro2 a1 a2 w11 w12 w22 b1 b2 b3 b11 b12 b13 b22 b23 b33 c11 c12
c21 c22 c13 c23 d1 d2 e1 e2 e3 f1 f2 if year==2002
matrix b0=e(b),ln(e(rmse)^2),.1

 matrix list b0

constraint drop _all

constraint define 1 [lnpro2]a1+[lnpro2]a2=1
constraint define 2 [lnpro2]w11+[lnpro2]w12+[lnpro2]w22=0
constraint define 3
[lnpro2]c11+[lnpro2]c21+[lnpro2]c12+[lnpro2]c22+[lnpro2]c13+[lnpro2]c23=0
frontier  lnpro2 a1 a2 w11 w12 w22 b1 b2 b3 b11 b12 b13 b22 b23 b33 c11
c12 c21 c22 c13 c23 d1 d2 e1 e2 e3 f1 f2 if year==2002, from (b0,copy)
iterate(1000) distribution(exponential) nolog constraints(1 2 3)

Ps: I am using a translog function.

*
*   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/