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

st: RE: Non-Linear Least Squares Programming Question

From   "Nick Cox" <>
To   <>
Subject   st: RE: Non-Linear Least Squares Programming Question
Date   Thu, 16 Aug 2007 19:01:18 +0100

You can't include full matrices on the LHS of 
a -replace- command. The syntax of -replace- is 

replace oldvar =exp [if] [in] [, nopromote]

and so what occurs on the LHS can _only_ be 
an existing variable name. 

The situation for the RHS is not quite so
clear-cut, but the whole of the RHS could not 
be, or evaluate to, a full matrix, if only for the reason above. 

In Stata 9, Mata or the matrix programming language
might be enough for your problem. 


Randall K. Q. Akee, Ph.D.
> I have a program that I want to use with the NL (non-linear least
> squares) command in Stata.  I just have a simple question 
> with regard to
> Stata v.9, for my objective function that gets "replaced" in the NL
> program, is it possible to have the Left Hand side and the ride hand
> side both be matrices?  Typically, I have seen examples (in the stata
> manual and other people's do files), where the right hand 
> side is simply
> a single equation.
> For instance:
> capture program drop nltest
> program define nltest    
>         version 9.0
> if "`1'"==  { global S_1 b1 b2 b3
>         }
>         replace `1'=$b1*X+exp($b2*$b3)
> end
> nl test data1 X etc. . .
> I want to know if Stata v.9 can handle a matrix on both the right hand
> side and left hand side (can data1 and X be matrices?) in the 
> line above
> that begins with "replace".  I have several equations that I 
> would like
> to fit with the same parameters.
> I know Stata v.10 has the NLSUR command, I just want to know 
> if I should
> wait for this version.

*   For searches and help try:

© Copyright 1996–2022 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index