[R-sig-eco] model selection with truncated poisson distribution

[James Rodger]

Hi Roman,

thanks for your response. I admit I was thinking of the 'dredging' type at
this stage. In this particular case, it probably won't make much difference
as my data doesn't show much in the way of patterns. I am going through the
model selection process to test for effects of treatment, plant density and
the interaction

cheers,

James

On Tue, Jan 6, 2015 at 2:00 PM, Roman Luštrik <roman.lustrik at gmail.com>
wrote:

> What kind of model selection are we talking about? The dredging part (a la
> step()) or a more fine tuned, such as information-theoretic approach
> advocated by (among others) Burnham and Anderson?
>
> I would suggest the latter. Pardon me if I'm presuming too much, but in a
> nutshell, this approach advocates formulating a set of plausible
> hypotheses, constructing models to reflect them, and then compare based on
> some criterion (like AIC). You can also use model averaging (using weights)
> to get a better sense of model estimates.
>
> Cheers,
> Roman
>
> On Tue, Jan 6, 2015 at 12:52 PM, James Rodger <rodgerjg at gmail.com> wrote:
>
>> Hi,
>>
>> Can anyone advise me on model selection in glm with a  zero truncated
>> poisson distribution?
>>
>> I have used a the command vglm in the vgam package to analyse the data
>> with
>> pospoisson specified as the family and I have evaluated alternative models
>> with a log-likelihood test. I am not  sure if that the test statistic will
>> follow the chi-squared distribution, however.
>>
>> My data is on seeds per fruit (minimum 1) in different experimental
>> treatments and different densities of plants.
>>
>> Regards,
>>
>> James
>>
>>         [[alternative HTML version deleted]]
>>
>> _______________________________________________
>> R-sig-ecology mailing list
>> R-sig-ecology at r-project.org
>> https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
>>
>
>
>
> --
> In God we trust, all others bring data.
>

	[[alternative HTML version deleted]]