Hi All,
Thanks for all of your help. I hadn't posted error messages, because I had
tried many different formulations of those functions and each one gave a
different error message. However, using your suggestions, I managed to make
it work.
You were right in that I only needed to call the constant parameters, c and
z. I wanted to define s within the nll() function. Mean and Var are
variables that are attached to the workspace in a dataframe. This worked:
Predict <- function(Mean,c,z)
{
v = c*Mean^z
return(v)
}
nll <- function(c,z){
n<-length(Mean)
s<-(sum(Var-Predict(Mean,c,z))^2)/n
logl<- -0.5*n*log(2*pi) -0.5*n*log(s) -
(1/(2*s))*sum((Var-Predict(Mean,c,z))^2)
return(-logl)
}
Mean <- rexp(130)
Var <- runif(130)
mle(nll,start=list(c=0.01,z=2.1))
It also worked for my actual data. It gives similar estimates for c and z as
nls, which is also a good sign, although, given normal errors, I believe nls
and maximum likelihood estimation should give the same parameter estimates.
Thanks again.
Anita.
On Thu, Jul 8, 2010 at 6:48 AM, Ravi Varadhan <rvarad...@jhmi.edu> wrote:
> Hi,
>
> When you report errors please send a reproducible example that will allow
> us to better help you. At a minimum, you should cut and paste the actual
> call and the error message.
>
> I think the problem is that you should only have parameters as arguments,
> i.e. only c and z should be in the argument list to nll().
>
> This works:
>
> Predict <- function(M,c,z){
> v = c*M^z
> return(v)
> }
>
> nll <- function(c,z){
> n<-length(M)
> logl<- -.5*n*log(2*pi) -.5*n*log(s) - (1/(2*s))*sum((V-Predict(M,c,z))^2)
> return(-logl)
> }
>
> M <- rexp(130)
> V <- runif(130)
> s <- 200
>
> mle(nll,start=list(c=0.01,z=2.1))
>
> > mle(nll,start=list(c=0.01,z=2.1))
>
> Call:
> mle(minuslogl = nll, start = list(c = 0.01, z = 2.1))
>
> Coefficients:
> c z
> 0.48547691 -0.01018601
> >
>
> Hope this helps,
> Ravi.
> ____________________________________________________________________
>
> Ravi Varadhan, Ph.D.
> Assistant Professor,
> Division of Geriatric Medicine and Gerontology
> School of Medicine
> Johns Hopkins University
>
> Ph. (410) 502-2619
> email: rvarad...@jhmi.edu
>
>
> ----- Original Message -----
> From: Anita Narwani <anitanarw...@gmail.com>
> Date: Wednesday, July 7, 2010 9:11 pm
> Subject: [R] Using nlm or optim
> To: r-help@r-project.org
>
>
> > Hello,
> > I am trying to use nlm to estimate the parameters that minimize the
> > following function:
> >
> > Predict<-function(M,c,z){
> > + v = c*M^z
> > + return(v)
> > + }
> >
> > M is a variable and c and z are parameters to be estimated.
> >
> > I then write the negative loglikelihood function assuming normal errors:
> >
> > nll<-function(M,V,c,z,s){
> > n<-length(Mean)
> > logl<- -.5*n*log(2*pi) -.5*n*log(s) -
> (1/(2*s))*sum((V-Predict(Mean,c,z))^2)
> > return(-logl)
> > }
> >
> > When I put the Mean and Variance (variables with 136 observations)
> > into this
> > function, and estimates for c,z, and s, it outputs the estimate for the
> > normal negative loglikelihood given the data, so I know that this works.
> >
> > However, I am unable to use mle to estimate the parameters c, z, and
> > s. I do
> > not know how or where the data i.e. Mean (M) and Variance (V) should
> > enter
> > into the mle function. I have tried variations on
> >
> > mle(nll,start=list(c=0.01,z=2.1,s=200)) including
> > mle(nll,start=list(M=Mean,V=Var, c=0.01,z=2.1,s=200))
> >
> > I keep getting errors and am quite certain that I just have a syntax
> > error
> > in the script because I don't know how to enter the variables into MLE.
> >
> > Thanks for your help,
> > Anita.
> >
> > [[alternative HTML version deleted]]
> >
> > ______________________________________________
> > R-help@r-project.org mailing list
> >
> > PLEASE do read the posting guide
> > and provide commented, minimal, self-contained, reproducible code.
>
[[alternative HTML version deleted]]
______________________________________________
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.
