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

st: RE: nonlinear least squares

From   "FEIVESON, ALAN H. (AL) (JSC-SK) (NASA)" <[email protected]>
To   "'[email protected]'" <[email protected]>
Subject   st: RE: nonlinear least squares
Date   Wed, 12 Jan 2005 08:59:52 -0600

Federica - I believe with a starting value of $G near zero, the program
seems to have trouble calculating numerical partial derivatives, hence it
won't ever change $G. 

I did some simulation with your model and sometimes using $G=1 as a starting
value would lead to reasonable results, but not always. The basic problem is
that there there is not enough sensitivity (to the value of G) in the
logarithmic term to allow efficient estimation of $G. 

To help things, I tried G=1 as a starting value and then did a simple linear
regression (with G = 1) to get initial values for $A, $B, etc. This can be
done in the initialization section of your nl-program. 

Al Feiveson

-----Original Message-----
From: [email protected]
[mailto:[email protected]]On Behalf Of Federica
Sent: Wednesday, January 12, 2005 6:34 AM
To: [email protected]
Subject: st: nonlinear least squares

Dear Stata users,

I am trying to estimate the following function by
nonlinear least squares:

Lnoutput = $A+$B*lnlabourhours + $D*lnmaterials +
$T*OtherCapital + $X*ln(Inv2003 + Inv2002/(1+$G) +
Inv2001/(1+$G)^2 + Inv2000/(1+$G)^3 + ...)

After running a do file to define the programme nltest
and set initial values for the parameters, I estimate
the model with these results:
- Stata takes 'parameter G as constant term in model'.
Also, it does not compute standard error, t statistic
- The value of G is 243,827, while the starting value
was 0.03!

It would be great if someone could help me understand
why this happens.

Thanks in advance for any help.

Nuovo Yahoo! Messenger: E' molto pi� divertente: Audibles, Avatar, Webcam,
Giochi, Rubrica... Scaricalo ora!
*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index