[R] bbmle "Warning: optimization did not converge"

[Uwe Ligges]

On 25.11.2012 19:44, arun4 wrote:
> Thank you  Michael Weylandt.
> Let me to describe my problem fully,
> I have developed a new discrete probability distribution which has the
> following Probability mas function( as an alternative to binomial
> distribution)
> <http://r.789695.n4.nabble.com/file/n4650759/newdist.jpg>
>
> Where n= number of trials
> x are the binomial values
> a and b are the two parameters to be estimated (I use MLE method by calling
> mle2 from bbmle package)
> As you can see, there is an inner summation which runs from zero to (n-x)
>
> The below is the R functions I have written to define Negative Loglikelihood
> and estimate parameters:
>
> library(bbmle)
>
> * ###Define Negative LL
> Dist.NLL<-function(x,a,b,fre,n) {
>    term<-0
>    for (j in 0:(n-x)) {
>      term=term+(((-1)**j)*(choose(n-x,j))*(beta(((x/a)+1+(j/a)),b)))
>    }
>    density=b*choose(n,x)*term
>    LL<-sum(fre*log(density))
>    return(-LL)
> }
>
> ##an example dataset
> x.values<-0:7  ##x values (here 7 trials)
> frequency<-c(47,54,43,40,40,41,39,95)   ##Observed frequencies of x values
> ##Now use mle2 to estimate parameters.
>
> mle2(Dist.NLL, start=list(a=22,b=22), data=list(x=values ,fre=frequency,
> n=7))
> *
>
> This is what I have done, bow I am getting "In 0:(n - x) : numerical
> expression has 8 elements: only the first used" error messages, which I
> afraid serious errors.

Dist.NLL obviously does not work for a vector of x values. You have to 
vectorize it.

Uwe Ligges




>
> Thanks again.
>
>
>
> --
> View this message in context: http://r.789695.n4.nabble.com/bbmle-Warning-optimization-did-not-converge-tp4650730p4650759.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> R-help at 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.
>