[R] Which distribution to select (massive fitting)

[David Winsemius]

On Feb 10, 2014, at 6:34 AM, Bert Gunter wrote:

> I believe this is more a question for SO (stats.stackexchange.com).

Actually it might get closed on SO since it is not an R programming question per se but rather an advice for statistical approach. It's a better fit for CrossValidated.com


> There are many possible goodness of fit statistics that can easily be
> calculated in R, but I think the fundamental question is: To what end?
> First, there are probably several parametric distributions that give
> (essentially) equally good fits; and second, you may want none of
> them, preferring some sort of nonparametric fit. Again, the sort of
> thing that is probably better at SO -- or even better, with a local
> statistician.


> Cheers,
> Bert
> 
> Bert Gunter
> Genentech Nonclinical Biostatistics
> (650) 467-7374
> 
> "Data is not information. Information is not knowledge. And knowledge
> is certainly not wisdom."
> H. Gilbert Welch
> 
> 
> 
> 
> On Mon, Feb 10, 2014 at 12:25 AM, Alaios <alaios at yahoo.com> wrote:
>> Hi all,
>> I have a large number of measurements from which I select a large number of unique vectors. For each vectors I would like to test which distribution might be a candidate for fitting.
>> It is impossible to look on each vector separately but I can inside a for loop test different models and based on their goodness of fit to make offline decisions (I will be saving goodness of fits results on a text file).
>> 
>> Do you know given a vector how I can get the goodness of fit for the "basic" distributions : "norm", "lnorm", "pois", "exp", "gamma", "nbinom",
>> "geom", "beta", "unif" and "logis"
>> 
>> Is it possible to try many of those (or at least some of the above) and try to get these results?
>> 
>> Regards
>> Alex
>>        [[alternative HTML version deleted]]
> 


David Winsemius
Alameda, CA, USA