# [R] Function for Distribution Fitting

Ebert,Timothy Aaron tebert @end|ng |rom u||@edu
Wed Oct 26 19:15:29 CEST 2022

```If you have three values, then it should be possible to fit any number of distributions such that they could return those three values. On the other hand, if you have billions of values and insist on a perfect fit then you may have a unique distribution only identified by the data in hand. For most purposes it is better to define the distribution based on your understanding of the numbers and how they were generated rather than trying to identify the distribution that best matches your data. I do not know your application for this. Here is a list of available functions in R:
https://cran.r-project.org/web/views/Distributions.html

I hope that helps, but I am not sure this is a task worth doing. If you start at the top of the list, and stop at the first "match" that sounds a bit like p-hacking.

Tim

-----Original Message-----
From: R-help <r-help-bounces using r-project.org> On Behalf Of JRG via R-help
Sent: Wednesday, October 26, 2022 10:01 AM
To: r-help using r-project.org
Subject: Re: [R] Function for Distribution Fitting

[External Email]

"all possible probability distributions"

I doubt that is a finite set.  To select "a bunch of them" would seem to imply a reduction of that set based on what's possible/promising/pertinent in your specific problem.  IMHO, what's the "most suitable distribution" tends to depend partly on knowledge of the subject matter of your problem.

Doesn't that put you back in your present situation?

---JRG

On 10/26/22 09:41, Paul Bernal wrote:
> Dear friends from the R community,
>
> Hope you are all doing great. So far, whenever I need to perform
> distribution fitting on a particular dataset, I make use of R package
> fitdistrplus.
>
> However, distribution fitting using the fitdist() function from
> fitdistrplus is rather manual (you need to specify which probability
> distribution you are trying to fit), and it would be more convenient
> if the function tested all possible probability distributions (or a
> bunch of them) and then estimates the parameters and suggests the most
> suitable distribution.
>
> Is there any package besides fitdistrplus that does allow automatic
> distribution fitting?
>
> Best regards,
> Paul
>
>       [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> .ethz.ch%2Fmailman%2Flistinfo%2Fr-help&data=05%7C01%7Ctebert%40ufl
> .edu%7Cf7aa511782ff4d98a04908dab75aa2a0%7C0d4da0f84a314d76ace60a62331e
> 1b84%7C0%7C0%7C638023897166448945%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4w
> LjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C
> &sdata=XS069Wk8JxeT4piBqA5JC3vtE2vu6uS8KTp1tqDIf7c%3D&reserved
> -project.org%2Fposting-guide.html&data=05%7C01%7Ctebert%40ufl.edu%
> 7Cf7aa511782ff4d98a04908dab75aa2a0%7C0d4da0f84a314d76ace60a62331e1b84%
> 7C0%7C0%7C638023897166448945%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwM
> DAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&
> sdata=FZHNApQa2raJnbAgUZ544bH91EMtjSG7Ef47dMCoo3M%3D&reserved=0
> and provide commented, minimal, self-contained, reproducible code.

______________________________________________
R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see