[R] Help in customising the NLS function to spit out mean and SD ofnew fit!!
Liaw, Andy
andy_liaw at merck.com
Fri Jan 7 22:03:21 CET 2005
Doesn't look like nonparametric fit to me, since nls() is used to fit to a
gaussian density, so the result is a gaussian density (with estimated
parameter).
What I do not understand is why people would do this. This is not the first
time I've seen people doing this, on both R-help and S-news (if my memory is
still any good). If the objective is to fit a Gaussian distribution to the
data, the `best' Gaussian curve is the one where mu = mean(data) and sigma^2
= var(data). Why go through all that trouble? What am I missing?
Andy
> From: Berton Gunter
>
> It **sounds ** like you are trying to fit a nonparametric
> density to 6000
> values... If so, please stop what you are doing and see
> ?density. You could
> also search on "fit density" or something similar on the R
> site search, as
> there are other R functions in R packages that do density
> fitting (ash is
> one, but other recommendations anyone?).
>
> If this is not what you are trying to do, I think you are
> still likely going
> about it in the wrong way -- histograms lose information and
> are notoriously
> dependent on the choice of cutpoints (which is part of the
> motivation for
> Scott's ash package). You might wish to consult a local
> statistician to get
> some better approaches to whatever it is that you're trying to do.
>
> Cheers,
> Bert Gunter
>
>
> -----Original Message-----
> From: r-help-bounces at stat.math.ethz.ch
> [mailto:r-help-bounces at stat.math.ethz.ch] On Behalf Of
> Jagarlamudi, Choudary
> Sent: Friday, January 07, 2005 10:50 AM
> To: r-help at stat.math.ethz.ch
> Subject: [R] Help in customising the NLS function to spit out
> mean and SD
> ofnew fit!!
>
> i'm coding in R(first time) for a paper my colleague is publishing.
> i plotted a histogram for 6000 values. Taking the mean and sd of the
> midpoints i did a dnorm and got the densities.
> pl<-dnorm(trimmedvals,mean=midsmean,sd=midsSD)
> (now in a loop of 5 times)
> i plotted these experimental values against theoretical values using
> the nls function the following way.
> nlsresult<-nlsModel(pl ~ 1/sqrt(2
> *pi)*std)*2.71828^2/(2*std4^2)),data=answer,start=list(std4=st
> d4,sm=sm))
> in the formula mean was mid point of 1st frequency bar and
> sd was sd of
> 6000 values.I do this 5 times each time changing the mean in
> the formula to
> be the mid point of the next frequency bar.
> now i plotted plot(fittedvals ~trimmedvals)
>
> I am told to plot experimental vs theoretical vlaues from the
> histogram and
> get a non linear least square curve fit.
> I need the mean and sd of this new fit to proceed to my next
> module.I'm
> not sure if i'm on track. Excuse me if this sounds too
> naive.If nls is not
> what i should be using can you please give me some pinters
> to solve this.
>
> Thanks in advance.
>
> Choudary Jagarlamudi
> Instructor
> Southwestern Oklahoma State University
> STF 254
> 100 campus Drive
> Weatherford OK 73096
> Tel 580-774-7136
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
> http://www.R-project.org/posting-guide.html
>
>
More information about the R-help
mailing list