Hi Kate and others,
thanks for the info.
Btw, you sent the different
methods to analyze the data: nls, nls.lm and nlrob. Comparing the
results visually nlrob performed better then nls, but nls.lm (using the
0.9 quantile of residuals) was still better than nlrob. My data may
have a rather large amo
Dear Martin,
Thanks for the ideas regarding the relation of what Fernando is doing with
robust regression. Indeed, it's an important point that he can't consider
the standard error estimates on his parameters correct.
I know from discussion off-list that he's happy with the results he has
now; n
Hi Kate and Fernando,
I'm late into this thread,
but from reading it I get the impression that Fernando really
wants to do *robust* (as opposed to least-squares) non-linear
model fitting. His proposal to set residuals to zero when they
are outside a given bound is a very special case of an
M-esti
You can take minpack.lm_1.1-0 (source code and MS Windows build,
respectively) from here:
http://www.nat.vu.nl/~kate/minpack.lm_1.1-0.tar.gz
http://www.nat.vu.nl/~kate/minpack.lm_1.1-0.zip
The bug that occurs when nprint = 0 is fixed. Also fixed is another
problem suggested your example: when th
You indeed found a bug. I can reproduce it (which I should have tried to
do on other examples in the first place!). Thanks for finding it.
It will be fixed in version 1.1-0 which I will submit to CRAN soon.
On Fri, 9 May 2008, elnano wrote:
>
> Find the data (data_nls.lm_moyano.txt) here:
> ft
Find the data (data_nls.lm_moyano.txt) here:
ftp://ftp.bgc-jena.mpg.de/pub/outgoing/fmoyano
Katharine Mullen wrote:
>
> Thanks for the details - it sounds like a bug. You can either send me the
> data in an email off-list or make it available on-line somewhere, so that
> I and other people ca
Thanks for the details - it sounds like a bug. You can either send me the
data in an email off-list or make it available on-line somewhere, so that
I and other people can download it.
On Fri, 9 May 2008, elnano wrote:
>
> Thank you Katharine. I am certain nprint is affecting my solution. Let me
Thank you Katharine. I am certain nprint is affecting my solution. Let me
know how I can send the data (~300Kb). The script I used it:
ST1 <- ST04
SM1 <- SM08
SR1 <- SRch2
ST <- ST1 [!is.na(SR1)]
SM <- SM1 [!is.na(SR1)]
SR <- SR1 [!is.na(SR1)]
q <- 0.90
p <- c("a"=-0.003, "b"=0.1
ct.org
> Sent: Thursday, May 8, 2008 5:43:31 PM
> Subject: Re: [R] function in nls argument
>
>
> I've basically solved the problem using the nls.lm function from the
> minpack.lm (thanks Katharine) with some modifications for ignoring residuals
> above a given percentil
8 5:43:31 PM
Subject: Re: [R] function in nls argument
I've basically solved the problem using the nls.lm function from the
minpack.lm (thanks Katharine) with some modifications for ignoring residuals
above a given percentile. This is to avoid the strong influence of points
which push my
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
I've basically solved the problem using the nls.lm function from the
minpack.lm (thanks Katharine) with some modifications for ignoring residuals
above a given percentile. This is to avoid the strong influence of points
which push my modeled vs. measured values away from the 1:1 line.
I based it o
The error message means that the gradient (first derivative of residual
vector with respect to the parameter vector) is not possible to work with;
calling the function qr on the gradient multiplied by the square root of
the weight vector .swts (in your case all 1's) fails.
If you want concrete adv
13 matches
Mail list logo
