Statalist The Stata Listserver

[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: Coelli

From   Annika Frohloff <>
Subject   st: Coelli
Date   Thu, 09 Nov 2006 14:34:59 +0100

Dear George,

which approach do you mean? Coelli 92 and 95 both assume a truncated normal distribution of the inefficiency term. The model I estimate incorporating exogenous variables z_i is:
frontier lny lnx1 ....lnxN, cm(Z1...ZM) d(t)
where cm stands for conditional mean model in Stata.

The error message could first mean, that the model is inappropriate for the data. Practically, this means that you should try to eliminate outliers from your data. Second try different numbers of efficiency variables (from none to all).
Check if the model converges in the simple half normal or exponential case without efficiency variables. If the simple case does not converge you should reconsider the input and output variables you plug into the model.

Another problem I faced was that in Stata the underlying solution algorithm is based on the method of moments to gain starting values which makes estimation more complicated.
What I do is to produce OLS starting values and plug those into the frontier estimation equation.
reg lny lnx1 ....lnxN
matrix b0 = e(b), ln(e(rmse)^2) , 0,...,0,.1 ,.1
/*where you put as many zeros (or different starting values) as there are efficiency variables*/
matrix list b0
frontier lny lnx1 ....lnxN, cm(Z1...ZM) d(t) from(b0, copy)

I hope that one of my ideas works.

Much luck, best,


george owuor schrieb:

I am trying to run production function using stochastic frontier following Coellis approach, but i get information in stata that reads: 'initial values not feasible'.

My expression reads like: frontier lnY lnX1------------------lnXn, distribution(exponential) uhet(efficiency variables: Z1--------Zn)

What could be the problem withmy expression. Please assist.


* For searches and help try:

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index