Not sure if this helps, but what error distribution are you using? ML is going to need a likelihood which requires you to pick an error model, but you only have a mean structure.
-----Original Message-----
From: "Magnus Soderberg" <[email protected]>
To: "[email protected]" <[email protected]>
Sent: 3/3/2009 5:39 PM
Subject: st: Nonlinear model using maximum likelihood
Dear All,
I want to estimate the following nonlinear function:
Y = b0 + b1*exp(b2*x1)*b4*x2*x3 + Xb
Where b0 to b3 are parameters (b vector of parameters), x1 to x3 variables (X a vector of variables).
I could estimate this by using the Stata command nl, but I want to use ML.
There is a similar problem and a solution posted at http://www.stata.com/support/faqs/stat/nl_ml.html which looks like
program mlnexpgr
version 10
args lnf b1x b0 sigma
tempvar res
quietly gen double `res' = $ML_y1 - `b0'*(1-exp(-`b1x'))
quietly replace `lnf' = -0.5*ln(2*_pi)-ln(`sigma')-0.5*`res'^2/`sigma'^2
end
ml model lf mlnexpgr (b1: rep78 = headroom, nocons) (b0:) (sigma:)
ml max
I guess my problem is the specification of the "ml model"-line but despite numerous attempts I can't get a reasonable output.
Does anyone know how to do this?
All the best,
Magnus
*
* 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/