R help - Nov 2005 - Distribution fitting problem

I am using the MASS library function 

fitdistr(x, dpois, list(lambda=2))

but I get 

Error in optim(start, mylogfn, x = x, hessian = TRUE, ...) :
        Function cannot be evaluated at initial parameters
In addition: There were 50 or more warnings (use warnings() to see the first 
50)

and all the first 50 warnings say 

1:  non-integer x = 1.452222
etc

Can anyone tell me what I am doing wrong. p.s. the data was read in from 
a .csv file that I wrote using octave

Mark Miller <mmiller <at> nassp.uct.ac.za> writes:
> 
> I am using the MASS library function 
> 
> fitdistr(x, dpois, list(lambda=2))
> 
> but I get 
> 
> Error in optim(start, mylogfn, x = x, hessian = TRUE, ...) :
>         Function cannot be evaluated at initial parameters
> In addition: There were 50 or more warnings (use warnings() to see the
first
> 50)
> 
The docs say:

For the following named distributions, reasonable starting values will be 
computed if start is omitted or only partially specified: cauchy, gamma, 
logistic, negative binomial (parametrized by mu and size), t and weibull. 

dpois is not among them, so you probably have to provide reasonable starting 
values for the parameters.

Dieter

Can you advise another distribution, was thinking of exponential, but was 
advised poisson since independent, forgot about requiring integers


On Wednesday 02 November 2005 14:44, you wrote:
> Mark Miller wrote:
> > I am using the MASS library function
> >
> > fitdistr(x, dpois, list(lambda=2))
> >
> > but I get
> >
> > Error in optim(start, mylogfn, x = x, hessian = TRUE, ...) :
> >         Function cannot be evaluated at initial parameters
> > In addition: There were 50 or more warnings (use warnings() to see the
> > first 50)
> >
> > and all the first 50 warnings say
> >
> > 1:  non-integer x = 1.452222
> > etc
> >
> > Can anyone tell me what I am doing wrong. p.s. the data was read in
from
> > a .csv file that I wrote using octave
>
> Hi, Mark,
>
> If you think the data are poisson, the observations should be integers.
>
> --sundar

On Wed, 2 Nov 2005 14:32:52 +0200, Mark Miller wrote:

MM> I am using the MASS library function 
MM> 
MM> fitdistr(x, dpois, list(lambda=2))
MM> 
MM> but I get 
MM> 
MM> Error in optim(start, mylogfn, x = x, hessian = TRUE, ...) :
MM>         Function cannot be evaluated at initial parameters
MM> In addition: There were 50 or more warnings (use warnings() to see
MM> the first  50)
MM> 
MM> and all the first 50 warnings say 
MM> 
MM> 1:  non-integer x = 1.452222
MM> etc
MM> 

are the data integers (as implicit in the assumption of Poisson dist'n)?
the above message seems to say that they are not 

Adelchi Azzalini

-- 
Adelchi Azzalini  <azzalini at stat.unipd.it>
Dipart.Scienze Statistiche, Universit?? di Padova, Italia
tel. +39 049 8274147,  http://azzalini.stat.unipd.it/

Hi, Mark,

Not without seeing you data. You only provide the first value is a 
warning message below.

--sundar

Mark Miller wrote:
> Can you advise another distribution, was thinking of exponential, but was 
> advised poisson since independent, forgot about requiring integers
> 
> 
> On Wednesday 02 November 2005 14:44, you wrote:
> 
>>Mark Miller wrote:
>>
>>>I am using the MASS library function
>>>
>>>fitdistr(x, dpois, list(lambda=2))
>>>
>>>but I get
>>>
>>>Error in optim(start, mylogfn, x = x, hessian = TRUE, ...) :
>>>        Function cannot be evaluated at initial parameters
>>>In addition: There were 50 or more warnings (use warnings() to see
the
>>>first 50)
>>>
>>>and all the first 50 warnings say
>>>
>>>1:  non-integer x = 1.452222
>>>etc
>>>
>>>Can anyone tell me what I am doing wrong. p.s. the data was read in
from
>>>a .csv file that I wrote using octave
>>
>>Hi, Mark,
>>
>>If you think the data are poisson, the observations should be integers.
>>
>>--sundar

At 10:32 2/11/2005, you wrote:
>I am using the MASS library function
>
>fitdistr(x, dpois, list(lambda=2))
>
>but I get
>
>Error in optim(start, mylogfn, x = x, hessian = TRUE, ...) :
>         Function cannot be evaluated at initial parameters
>In addition: There were 50 or more warnings (use warnings() to see the first
>50)
>
>and all the first 50 warnings say
>
>1:  non-integer x = 1.452222
>etc
>
>Can anyone tell me what I am doing wrong. p.s. the data was read in from
>a .csv file that I wrote using octave
Mark,

Try fitdistr(x, "Poisson")

I think this is enough for fit Poisson distribuition for your data


Bernardo Rangel Tura, MD, MSc
National Institute of Cardiology Laranjeiras
Rio de Janeiro Brazil 


--