On Mon, 15 May 2006, Pair Pierre-Matthieu wrote:
> Hello,
>
> I am currently writing a program whose goal is to fit usual
> distributions (estimating parameters and confidence intervals for a
> given distribution).
>
> After some research in R, R-help and google I have found most of what I
> was looking for (especially thanks to MASS - fitdistr() ), however there
> are still a few distributions I could not find R code for: Multinormal,
> Truncated normal, Triangular, Uniform, Binomial, Multinomial.
>
> If there are any packages to fit these, a pointer in the right direction
> would be most appreciated.
You don't need R code for most of these.
Binomial, Multinomial: the MLEs are the sample proportions, and the se's
are textbook formulae. Can also be done in fitdistr.
Triangular, Uniform: non-standard estimation problems, but the MLEs of the
support are the sample maximum and minimum. For a CI, you will need a
profile likelihood interval, or some such.
Truncated normal: pretty easy to do in fitdistr, provided the truncation
point is known (is it?).
Multinormal: the MLEs are the sample mean and the sample covariance
(divisor n).
--
Brian D. Ripley, ripley at stats.ox.ac.uk
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272866 (PA)
Oxford OX1 3TG, UK Fax: +44 1865 272595