Hey Henrik I dont do MLE myself but this recent blog might be helpful.
http://www.johnmyleswhite.com/notebook/2010/04/21/doing-maximum-likelihood-estimation-by-hand-in-r/ -A On Wed, Apr 21, 2010 at 10:02 AM, Thomas Stewart <tgstew...@gmail.com> wrote: > Two possible problems: > > (a) If you're working with a normal likelihood---and it seems that you > are---the exponent should be squared. As in: > ... + (1/(2*stdev*stdev))*sum( ( y-(rev/12)-lag(y)*exp(-lap/12) )^2 ) > > (b) lag may not be working like you think it should. Consider this silly > example: > y<-c(1,2,3,5,34,2,3,5,2) > lag(y)-y > lag(ts(y))-ts(y) > > Good luck, > -tgs > > > > On Wed, Apr 21, 2010 at 8:14 AM, Henkep <flyerhe...@hotmail.com> wrote: > >> >> Dear R-Help, >> >> I also send the following post by e-mail to you, however I try to post it >> here aswell. My name is Henrik and I am currently trying to solve a Maximum >> Likelihood optimization problem in R. Below you can find the output from R, >> when I use the "BFGS" method: >> >> The problem is that the parameters that I get are very unreasonable, I >> would >> expect the absolute value of each parameter to be bounded by say 5. >> (furthermore the variable stdev should be greater than zero). >> >> One of the problems seems to be that I need to bound the stdev-variable >> from >> below by zero to avoid the NaN:s produced. I unfortunately do not know how >> to do that. >> >> Below "y" is the dataset, to which, I want to fit the parameters. I.e. y is >> the vector (or R equivalent) of observations. >> I have also multiplied the log-likelihood function by -1, since I know that >> R by default minimizes the objective function. >> >> I would be very happy if you can come up with some Ideas on what is going >> wrong in the code below. >> >> Thank you for your time! >> >> Henrik >> >> >> > data<-read.table(file="c:/users/oukal.csv",header=TRUE) >> > y<-data$V1 >> > loglik<-function(estimate,y) >> + { >> + lap<-estimate[1] >> + stdev<-estimate[2] >> + rev<-estimate[3] >> + n<-length(y) >> + >> + 0.5*n*log(2*pi)+ 0.5*n*log(stdev) + >> + (1/(2*stdev*stdev))*sum(y-(rev/12)-lag(y)*exp(-lap/12)) >> + } >> > >> > optim(c(4,4,4),loglik,y=y,method="BFGS") >> $par >> [1] -4884.34155 445.52350 88.53777 >> $value >> [1] -1.910321e+174 >> $counts >> function gradient >> 290 7 >> $convergence >> [1] 0 >> $message >> NULL >> There were 50 or more warnings (use warnings() to see the first 50) >> >> -- >> View this message in context: >> http://n4.nabble.com/Maximum-Likelihood-Estimation-in-R-tp2018822p2018822.html >> Sent from the R help mailing list archive at Nabble.com. >> >> ______________________________________________ >> 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. >> > > [[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. > ______________________________________________ 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.
