s-news
[Top] [All Lists]

nlregb difficulty

To: s-news@wubios.wustl.edu
Subject: nlregb difficulty
From: David Parkhurst <parkhurs@imap.iu.edu>
Date: Thu, 10 May 2007 23:00:37 -0400
User-agent: Thunderbird 1.5.0.10 (Windows/20070221)
I'm using S-PLUS Enterprise Developer Version 7.0.6 for Microsoft Windows, and want to run a constrained non-linear regression. The help page for nlregb oddly provides only an example of *linear* regression, and I don't see just how to use the function for my problem. Here's what I've tried, with a trial dataset:

tC <- c(rep(0,5),rep(1,5)) # Dummy var for cooling cycle
tH <- c(rep(1,5),rep(0,5)) # Dummy var for warming cycle
restime <- rnorm(10,6,0.5)
fake <- data.frame(tC,tH,restime)

n <- length(fake$restime)

resid.func <- function(x,fake)
{-(1/x[1])*
  (fake$tC * log((x[2]-2)/(x[3]-2)) +
   fake$tH * log((x[3]-38)/(x[2]-38))) -
   fake$restime }

NLLS.constrained <- nlregb(
   nres=n,
   start=c(0.004,4,36),   # parms: k, m, M
   residuals=resid.func,
   lower=c(0,2+0.1,2+0.1),
   upper=c(Inf,38-0.1,38-0.1),
   data=fake)

I get this error message:
Problem in .Fortran("n2b0",: subroutine n2b0: Argument 6 has zero length
Use traceback() to see the call stack

I'm guessing that the parameters aren't getting to the residual function
correctly, but I don't see how to get them there.  I'll appreciate any
help.  Thanks.

David Parkhurst

<Prev in Thread] Current Thread [Next in Thread>
  • nlregb difficulty, David Parkhurst <=