Kia ora
I'm a using lme (from nlme package) with data similar to the Orthodont
dataset and am getting positive log-likelihoods (>100). This seems usual and
I wondered if someone could offer a possible explanation.
I can supply a sample dataset if requested, but I feel almost certain that this
question has been asked and answered recently. However, I can find no trace of
it in the mail archives (although I have spent several hours reading lots of
other interesting things :-)).
Thanks .........
Peter Alspach
______________________________________________________
The contents of this e-mail are privileged and/or confidenti...{{dropped}}
The "likelihood" is the probability density function, which can be
greater than 1 for continuous distributions with a fairly narrow
spread. For discrete distributions, the density never exceeds 1, in
which case the log(likelihood) would always be negative(*).
hope this helps.
spencer graves
(*) If you are using measure-theoretic probability with some
non-standard measure, it might be possible to get a discrete probability
density greater than 1. One might want to use such as a class exercise,
but I can't think of a real world application for such.
Peter Alspach wrote:
>Kia ora
>
>I'm a using lme (from nlme package) with data similar to the Orthodont
dataset and am getting positive log-likelihoods (>100). This seems usual and
I wondered if someone could offer a possible explanation.
>
>I can supply a sample dataset if requested, but I feel almost certain that
this question has been asked and answered recently. However, I can find no
trace of it in the mail archives (although I have spent several hours reading
lots of other interesting things :-)).
>
>Thanks .........
>
>Peter Alspach
>
>
>______________________________________________________
>
>The contents of this e-mail are privileged and/or confidenti...{{dropped}}
>
>______________________________________________
>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
>
>
Hi Peter, Why do you think positive log-likelihoods are unusual? consider:> dnorm(1,1,0.1)[1] 3.989423> log(dnorm(1,1,0.1))[1] 1.383647 Any log-likelihood would be a sum of such terms. Hth, ingmar On 2/16/05 11:02 PM, "Peter Alspach" <PAlspach at hortresearch.co.nz> wrote:> > Kia ora > > I'm a using lme (from nlme package) with data similar to the Orthodont dataset > and am getting positive log-likelihoods (>100). This seems usual and I > wondered if someone could offer a possible explanation. > > I can supply a sample dataset if requested, but I feel almost certain that > this question has been asked and answered recently. However, I can find no > trace of it in the mail archives (although I have spent several hours reading > lots of other interesting things :-)). > > Thanks ......... > > Peter Alspach > > > ______________________________________________________ > > The contents of this e-mail are privileged and/or confidenti...{{dropped}} > > ______________________________________________ > 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-- Ingmar Visser Roetersstraat 15 1018 WB Amsterdam The Netherlands i.visser at uva.nl http://users.fmg.uva.nl/ivisser/