[R-SIG-Mac] problem with confint.nls on Intel Mac but not powerpc
kknoblauch at free.fr
kknoblauch at free.fr
Thu Nov 30 18:01:11 CET 2006
Simon,
Thank you for trying this out. I just updated to today's
patched (thank for making that into such an easy thing to
do), though I get "page not found" for the nightly build
of the Rgui.app (that will certainly be accessible later).
So, I've retried and I get the same error. So, it must
be something on my system and not a an Intel vs powerpc
problem, since it works for you. That hypothesis is out
the window.
I"m running Mac OS X 10.4.8, 2.165 Ghz Intel Core Duo,
2 GB 667 MHz DDR2 SDRAM, in case that matters.
Since I can solve the problem on the Intel by adding
derivative information to the model (and also by running
it on my powerpc), I'm not suffering from not having a
solution. As you demonstrate, however, it should work.
Can you suggest where I might look? Thanks.
best,
ken
R version 2.4.0 Patched (2006-11-29 r40065)
i386-apple-darwin8.8.1
locale:
C
attached base packages:
[1] "stats" "graphics" "grDevices" "utils" "datasets"
[6] "methods" "base"
> dd <- read.table("mesures_240906.txt", header = TRUE, dec = ",")
> dd$Gun <- factor(dd$Gun, levels = c("rouge", "vert", "bleu",
+ "gris"))
> dd <- dd[, c(2, 1, 3)]
> dd$GL <- dd$GL/255
> ddsub <- subset(dd, Gun != "gris")
> ddsub$Gun <- ddsub$Gun[, drop = TRUE]
> gg <- model.matrix(~-1 + Gun/GL, dd)[1:51, c(5:7)]
> dd.plin <- nls(Lum[1:51] ~ cbind(rep(1, 51), gg^gamm), data = ddsub,
+ start = list(gamm = 2.4),
+ alg = "plinear"
+ )
> B <- as.vector(coef(dd.plin))
> st <-list(Blev = B[2], beta = B[3:5], gamm = B[1])
> dd.nls <- nls(Lum ~ Blev + beta[Gun] * GL^gamm,
+ data = ddsub, start = st)
> summary(dd.nls)
Formula: Lum ~ Blev + beta[Gun] * GL^gamm
Parameters:
Estimate Std. Error t value Pr(>|t|)
Blev -0.06552 0.04766 -1.375 0.176
beta1 16.74186 0.13245 126.402 <2e-16 ***
beta2 50.51605 0.14802 341.281 <2e-16 ***
beta3 8.12063 0.13452 60.368 <2e-16 ***
gamm 2.51187 0.01561 160.916 <2e-16 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 0.2067 on 46 degrees of freedom
> confint(dd.nls)
Waiting for profiling to be done...
Error in prof$getProfile() : step factor 0.000488281 reduced below 'minFactor'
of 0.000976562
platform i386-apple-darwin8.8.1
arch i386
os darwin8.8.1
system i386, darwin8.8.1
status Patched
major 2
minor 4.0
year 2006
month 11
day 29
svn rev 40065
language R
version.string R version 2.4.0 Patched (2006-11-29 r40065)
Quoting Simon Urbanek <simon.urbanek at r-project.org>:
> Ken,
>
> I don't know really what it is, but your example works fine in both
> current R 2.4.0 patched and in R-devel even on my Intel iMac. Please
> test it as well and let me know if that solves your problem.
>
> Cheers,
> Simon
>
>
> On Nov 29, 2006, at 4:39 PM, Ken Knoblauch wrote:
>
> > I recently posted a question on one subject on the main R-help list
> > but was
> > stymied in finding a response because the error with confint.nls that
> > was occurring
> > on my Intel Mac could not be reproduced under linux.
> >
> > (https://stat.ethz.ch/pipermail/r-help/2006-November/117733.html)
> >
> > I finally came up with a solution to my initial problem that
> > satisfied my needs
> >
> > (https://stat.ethz.ch/pipermail/r-help/2006-November/117800.html)
> >
> > but I tried my problem just now on a powerpc, and I don't get the
> > error there.
> > So, the question is why is the Intel Mac producing this error?
> >
> > I paste below the example from the Intel Mac and then from the powerpc
> >
> > Intel Mac (this copied from the above listing on R-help)
> >
> > ####The data, again
> > dd
> > Lum GL Gun mBl
> > 0.15 0 rouge 0.09
> > 0.07 15 rouge 0.01
> > 0.1 31 rouge 0.04
> > 0.19 47 rouge 0.13
> > 0.4 63 rouge 0.34
> > 0.73 79 rouge 0.67
> > 1.2 95 rouge 1.14
> > 1.85 111 rouge 1.79
> > 2.91 127 rouge 2.85
> > 3.74 143 rouge 3.68
> > 5.08 159 rouge 5.02
> > 6.43 175 rouge 6.37
> > 8.06 191 rouge 8
> > 9.84 207 rouge 9.78
> > 12 223 rouge 11.94
> > 14.2 239 rouge 14.14
> > 16.6 255 rouge 16.54
> > 0.1 0 vert 0.04
> > 0.1 15 vert 0.04
> > 0.17 31 vert 0.11
> > 0.46 47 vert 0.4
> > 1.08 63 vert 1.02
> > 2.22 79 vert 2.16
> > 3.74 95 vert 3.68
> > 5.79 111 vert 5.73
> > 8.36 127 vert 8.3
> > 11.6 143 vert 11.54
> > 15.4 159 vert 15.34
> > 19.9 175 vert 19.84
> > 24.6 191 vert 24.54
> > 30.4 207 vert 30.34
> > 36.1 223 vert 36.04
> > 43 239 vert 42.94
> > 49.9 255 vert 49.84
> > 0.06 0 bleu 0
> > 0.06 15 bleu 0
> > 0.08 31 bleu 0.02
> > 0.13 47 bleu 0.07
> > 0.25 63 bleu 0.19
> > 0.43 79 bleu 0.37
> > 0.66 95 bleu 0.6
> > 1.02 111 bleu 0.96
> > 1.46 127 bleu 1.4
> > 1.93 143 bleu 1.87
> > 2.49 159 bleu 2.43
> > 3.2 175 bleu 3.14
> > 3.96 191 bleu 3.9
> > 4.9 207 bleu 4.84
> > 5.68 223 bleu 5.62
> > 6.71 239 bleu 6.65
> > 7.93 255 bleu 7.87
> >
> > ###For initial values - this time using plinear algorithm insted of
> > optim
> > gg <- model.matrix(~-1 + Gun/GL, dd)[ , c(4:6)]
> > dd.plin <- nls(Lum ~ cbind(rep(1, 51), gg^gamm), data = dd,
> > start = list(gamm = 2.4),
> > alg = "plinear"
> > )
> > B <- as.vector(coef(dd.plin))
> > st <-list(Blev = B[2], beta = B[3:5], gamm = B[1])
> >
> >
> > dd.nls <- nls(Lum ~ Blev + beta[Gun] * GL^gamm,
> > data = dd, start = st
> > )
> >
> > ### Here is the error that I observe
> > confint(dd.nls)
> > Waiting for profiling to be done...
> > Error in prof$getProfile() : step factor 0.000488281 reduced below
> > 'minFactor' of 0.000976562
> >
> > R version 2.4.0 Patched (2006-11-16 r39921)
> > i386-apple-darwin8.8.1
> >
> > locale:
> > C
> >
> > attached base packages:
> > [1] "stats" "graphics" "grDevices" "utils" "datasets"
> > [6] "methods" "base"
> >
> > other attached packages:
> > MASS lattice
> > "7.2-29" "0.14-13"
> > (still get this problem with today's patched version, however)
> >
> > and now from the powerpc, the relevant code
> >
> > st <-list(Blev = B[2], beta = B[3:5], gamm = B[1])
> > dd.nls <- nls(Lum ~ Blev + beta[Gun] * GL^gamm,
> > data = dd, start = st)
> > confint(dd.nls)
> > Waiting for profiling to be done...
> > 2.5% 97.5%
> > Blev -0.1612492 0.02988387
> > beta1 16.4752317 17.00843251
> > beta2 50.2193167 50.81325840
> > beta3 7.8496610 8.39135292
> > gamm 2.4811018 2.54296647
> >
> > R version 2.4.0 Patched (2006-11-24 r39989)
> > powerpc-apple-darwin8.8.0
> >
> > locale:
> > en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
> >
> > attached base packages:
> > [1] "stats" "graphics" "grDevices" "utils" "datasets"
> > "methods"
> > [7] "base"
> >
> > other attached packages:
> > MASS
> > "7.2-29"
> >
> > The Intel is having problems with some of the control
> > parameter settings for nls but I don't know why.
> > I hope that I have supplied sufficient information
> > for my question.
> > I would be thankful for any light that could be shed
> > on it.
> >
> >
> > ken
> >
> > --
> > Ken Knoblauch
> > Inserm U371
> > Institut Cellule Souche et Cerveau
> > Département Neurosciences Intégratives
> > 18 avenue du Doyen Lépine
> > 69500 Bron
> > France
> > http://www.lyon.inserm.fr/371/
> >
> >
> >
> >
> >
> >
> > [[alternative HTML version deleted]]
> >
> > _______________________________________________
> > R-SIG-Mac mailing list
> > R-SIG-Mac at stat.math.ethz.ch
> > https://stat.ethz.ch/mailman/listinfo/r-sig-mac
>
>
More information about the R-SIG-Mac
mailing list