nls factor

A more compact way to code factors in nls is to use the syntax factor[].

Here's an example using a simplified version of Ravi's example:

n <- 200
set.seed(123)
x <- runif(n)
a <- gl(n=2, k=n/2)  # a two-level factor

eps <- rnorm(n, sd=0.5)
y <- as.numeric(a) * x^.5 + eps

nls(y ~ a[]*x^b, start=list(a=c(1,1), b=0.25))

Wed, 2009-04-15 at 10:22 -0400, Ravi Varadhan wrote:

Hi,

Here is one-way to do it (the following code shows a simulation example):

n <- 200

set.seed(123)

x <- runif(n)

f <- gl(n=2, k=n/2)  # a two-level factor

x1 <- x * (f == 1)

x2 <- x * (f == 2)

a <- c(rep(2, n/2), rep(5, n/2))

b <- 0.5

nsim <- 100

nls.coef <- matrix(NA, nsim, 3)

for (i in 1:nsim) {

eps <- rnorm(n, sd=0.5)

y <- a * x^b + eps

ans.nls <- try(nls(y ~ a1*x1^b + a2*x2^b, start=list(a1=1, a2=1, b=0.25)),
silent=TRUE)

if (class(ans.nls) != "try-error") nls.coef[i, ] <- coef(ans.nls)

}

apply(nls.coef, 2, summary)  
 
Hope this helps,
Ravi.

----------------------------------------------------------------------------
-------

Ravi Varadhan, Ph.D.

Assistant Professor, The Center on Aging and Health

Division of Geriatric Medicine and Gerontology 

Johns Hopkins University

Ph: (410) 502-2619

Fax: (410) 614-9625

Email: rvaradhan at jhmi.edu

Webpage:  http://www.jhsph.edu/agingandhealth/People/Faculty/Varadhan.html

 

----------------------------------------------------------------------------
--------


-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On
Behalf Of Manuel Gutierrez
Sent: Wednesday, April 15, 2009 8:27 AM
To: r-help at r-project.org
Subject: [R] nls factor

I want to fit the model y=a*x^b using nls; where "a" should be different for
each level of a factor.
What is the easiest way to fit it? Can i do it with nls?
I've looked the help pages and the MASS example in page 249 but the formula
is different and I don't know how to specify it for my model.
Thanks,
Manuel

	[[alternative HTML version deleted]]