All the distribution tests are rule out tests, i.e. they can tell you
if your data does not match a given distribution, but they can never
tell you that the data does come from a specific distribution.
Note also that the results of any of these studies may not be that
useful, for small sample sizes it is more important to rule out a
given distribution, but unless there is a huge difference you won't
have much power to do this. For large sample sizes it is less
important because using a close distribution will generally give you
robust results, but you will have power to detect small, meaningless
differences. So often your choice is between a meaningless answer to
a meaningful question or a meaningful answer to a meaningless
question.
What is more important and a better approach is to understand the
science behind the process that generated the data and use that
knowledge to find a distribution that is reasonable (even if not
exact) or to use techniques that make fewer assumptions about the
distribution if you cannot find something close enough to be
reasonable (e.g. bootstrap, permutation, other non-parametric,
simulations to determine cut-off values).
On Tue, Feb 14, 2012 at 4:21 AM, Bianca A Santini
<b.santini at sheffield.ac.uk> wrote:> Hello!
> I have several variables. Each of them has a different distribution. I was
> thinking to use a Generalized Linear Model, glm(), but I need to introduce
> the family. Do you know if R has any tests for matching data to any
> distribution ( I am aware of shapiro.test).
>
> All the best,
>
>
> --
> BAS
>
> ? ? ? ?[[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
--
Gregory (Greg) L. Snow Ph.D.
538280 at gmail.com