Hi, All:
The following identifies apparent inconsistencies and proposed
fixes between the documentation and the behavior of 'nls'.
Specifically, the help file says 'data' 'Can also be a
list'.
However, I got an error when I tried it with a list that could not be
coerced to a data.frame. This is documented in the first example below.
Also, the help file says 'data' is 'optional'.
Unfortunately, the
second example below bombs, apparently while checking the wrong thing
for parameters to estimate.
Attached is a version of 'nls' that fixes these problems and
otherwise produces the same answers (per 'all.equal') as the current
official version on the "Examples" in 'nls' help file. What
do you
think about adopting these code changes and adding these examples to the
'nls' help file (see the attached 'nls-new examples.R')?
I am using this in preparing for a presentation on the 'fda'
package at UseR! in Ames (Aug. 8-10). I am temporarily adding the
attached version of 'nls' to the 'fda' package with a help file
primarily consisting of a pointer to help(nls, package=stats) and the
attached examples. However, I'm hoping that someone will find the time
to review this and either disabuse me of some of my deficiencies in this
area or announce a plan to have something like this incorporated in a
future version of R.
Thanks,
Best Wishes,
Spencer Graves
p.s. How can I find the official source code for this? I tried
"https://svn.r-project.org/R-packages/trunk" and found 'nlme'
and other
packages but not core R like 'stats'.
#####################
##
## SLIGHT MODIFICATIONS TO CURRENT EXAMPLES
## THAT DON'T WORK (except with my modifications)
##
# Weighted Michaelis-Menten model
# with data = a list that can not be coerced to a data.frame
TreatIrreg <- with(Treated, list(conc1=conc[1], conc.1=conc[-1], rate=rate))
# Passing arguments using a list that can not be coerced to a data.frame
weighted.MM1 <- function(resp, conc1, conc.1, Vm, K){
conc <- c(conc1, conc.1)
resp <- TI$rate
#
pred <- (Vm * conc)/(K + conc)
(resp - pred) / sqrt(pred)
}
Pur.wt1 <- nls( ~ weighted.MM1(rate, conc1, conc.1, Vm, K),
data=TreatIrreg,
start = list(Vm = 200, K = 0.1), trace = TRUE)
##
## ERROR MESSAGE:
## Error in data.frame(conc1 = 0.02, conc.1 = c(0.02, 0.06, 0.06, 0.11,
0.11, :
## arguments imply differing number of rows: 1, 11, 12
##
# Passing arguments via 'get'
weignted.MM0 <- function(Vm, K){
TI <- get("TreatIrreg")
conc <- with(TI, c(conc1, conc.1))
resp <- TI$rate
#
pred <- (Vm * conc)/(K + conc)
(resp - pred) / sqrt(pred)
}
Pur.wt0 <- nls( ~ weighted.MM0(Vm, K), start = list(Vm = 200, K = 0.1),
trace = TRUE)
##
## ERROR MESSAGE:
## Error in nls(~weighted.MM0(Vm, K), start = list(Vm = 200, K = 0.1),
trace = TRUE) :
## no parameters to fit
##
############################
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: nls.R
Url:
https://stat.ethz.ch/pipermail/r-devel/attachments/20070604/45e4d0d8/attachment.pl
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: nls-new examples.R
Url:
https://stat.ethz.ch/pipermail/r-devel/attachments/20070604/45e4d0d8/attachment-0001.pl
