Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: NLLS problem


From   "Brian P. Poi" <bpoi@stata.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: NLLS problem
Date   Sun, 25 Sep 2005 10:35:36 -0500 (CDT)

On Sat, 24 Sep 2005 btallaire@mail.utexas.edu wrote:

Stata users -
Hello, I am try to estimate a wage model by non-linear least squares.

wage = exp({educ}*educ+{exper}*exper+{espersq}*espersq+{smsa}*smsa+{south}*south)+ {delta}*(exp({educ2}*educ+{exper2}*exper+{espersq2}*espersq+{smsa2}*smsa+
{south2}*south)^2))

Every time I attempt to estimate it, stata gives me "
* Parameter south2 taken as constant term in model & ANOVA table"

Why is this? What am I doing incorrect? I've tried different starting values for south2 and it doesn't seem to work.

Any help would be most appreciated!
Thanks!!!
Ben

Ben

By default -nl- tries to identify a constant term in the model. It does that by computing the derivative d f(x;b) / d b_i for each parameter b_i. If the standard deviation of a derivative across all the observations in the estimation sample is less than a small fraction of the mean of the derivative [ e*(|mean| + tau), where tau = 0.001 and by default e=0.00001 to be precise], then that parameter is assumed to be the constant term in the model. If several parameters happen to meet that criterion, then the last one is treated as the constant term.

To see why that works, think about the linear regression y_i = b_0 + b_1 x_i + error_i. The derivative with respect to b_0 is one for all observations, so that the standard deviation is zero; and the criterion mentioned above is satisfied.

If you do not have a constant term in the model, and you do not want -nl- to search for one, use the "noconstant" option.

Hope this helps

-- Brian
-- bpoi@stata.com
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/




© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index