R help - Aug 2008 - Determining model parameters

This may be a begining question. If so, please bear with me. 

If I have some data that based on the historgram and other plots it
"looks" like a beta distribution. Is there a function or functions
within R to help me determine the model parameters for such a distirbution?
Similarily for other "common" distirbutions, Poisson(lambda),
Chi-Square(degrees of freedom, chi-square statistic), etc. For a normal
distribution it is relatively straight forward as mean and variance can be
computed failrly readily.
So far in my reading such parameters are more or less given but to do any kind
of inference using a prior model it seems key that these parameters be at least
estimated accurately.

Thank you.

Kevin

<rkevinburton <at> charter.net> writes:
> If I have some data that based on the historgram and other plots it
"looks"
like a beta distribution. Is there
> a function or functions within R to help me determine the model parameters
for
such a distirbution?


library(MASS)
?fitdistr

~  The beta distribution only applies to data that are bounded
between 0 and 1 (in some cases strictly bounded, i.e. 0<x<1 not
just 0<=x<=1).

~   As a start, try something like

d2 = (Diff-min(Diff)+0.001)/(max(Diff)-min(Diff)+0.002)
fitdistr(d2,densfun="beta",start=list(shape1=3,shape2=2))

~ [cc'ing back to R-help for posterity ...]

~ Ben Bolker

rkevinburton at charter.net wrote:
| OK. I see in the documentation that "start" is required for certain
distributions. Now I get:
|
| fitdistr(Diff, densfun="beta", list(shape1=3,shape2=2))
| Error in optim(x = c(0, 516, 968, 703, 960, 32, 956, 1417, -92, 471,
531,  :
|   initial value in 'vmmin' is not finite
|
| Any idea what vmmin is? And what the cause of the error is?
|
| Thanks again.
|
| Kevin
| ---- Ben Bolker <bolker at ufl.edu> wrote:
|>  <rkevinburton <at> charter.net> writes:
|>
|>> If I have some data that based on the historgram and other plots it
"looks"
|> like a beta distribution. Is there
|>> a function or functions within R to help me determine the model
parameters for
|> such a distirbution?
|>
|>
|> library(MASS)
|> ?fitdistr
|>