[Limdep Nlogit List] MLE codes for a simple SFA estimation

Dev Vencappa Dev.Vencappa at nottingham.ac.uk
Fri Jul 23 20:57:38 EST 2010


Dear Limdep users, 

 I am a fairly new user of Limdep, having started to use it only since
the start of the year. I am beginning to appreciate the many econometric
models it allows one to estimate through the standard built in commands
- Stochastic Frontier Analysis is one such example. I am embarking on
maximum likelihood estimation and have looked at the many examples on
this from the manual and other resources on the net. In the process, I
have been trying to replicate the estimation of a simple stochastic
production frontier by coding the non-linear optimisation process (for
the half-normal model) but this is not working out. Below is what I have
written - I guess the problem for me identifying properly to Limdep
which are the parameters to be estimated and within this estimation
clearly identify the scalars. I have read the chapter on Frontier Models
and Efficiency Analysis and came across the Olsen parameterization that
is used to estimate the half-normal model but have been unable to code
these, hence I went for the elaborate coding of the log likelihood
function. Any help towards improving these codes and getting them to
work would be much appreciated. I am attaching the data set I have used
so this can be checked. 

Many thanks in advance for your help.

Regards,
Dev


clear$
reset$
load;file="frontier1.lpj" $
Create; y=lnoutput;x1=lnlabor;x2=lncapita$
Namelist;X=one,x1,x2$

Regress; Lhs=y;Rhs=X $
MATRIX;BOLS=B;LIST $
Maximize
;start=BOLS,1,1,1
;labels=b0,b1,b2,su,sv,sig
;Fcn=d=.5*Log(2/Phi)-log(sig)+0.5*(((y-b0-b1*x1-b2*x2)/sig)^2)+log(N01((
y-b0-b1*x1-b2*x2)*(su/sv)/sig))
alg=nr; tlf=0.001; output=3$

  Error   759: Name in expr not label,variable,scalar,matrix,list,or
FNi.
  Error   588: Error translating function in optimization command.
  Error   589: Error is in 2/PHI
  Error    69: WALD,NLSQ,MAXIMIZE. Cannot translate the listed string

This message has been checked for viruses but the contents of an attachment
may still contain software viruses which could damage your computer system:
you are advised to perform your own checks. Email communications with the
University of Nottingham may be monitored as permitted by UK legislation.


More information about the Limdep mailing list