[R] p-values

[John Maindonald]

Frank -
Thanks for your reply, on which I really have no comment.

There are contexts where a Bayesian approach necessary,
natural and easy to handle, and can be used to broaden
the inferential vision of students.  Examples from HIV
testing and mammography screening in
Gigerenzer's book, sold in the USA under the title:
"Calculated Risks: How to Know When Numbers Deceive You"
(Simon & Schuster)
can be used to make highly important points.
I have, in the section on inference in my book with John Braun,
injected some of this into the discussion of inference.

[Incidentally, Frank's book does have a great deal of
practically oriented comment (e.g., wrt variable selection)
that I have found useful. There is useful critical comment
on inappropriate use of p-values in such contexts.]

John Maindonald.

On 3 May 2004, at 9:58 AM, Frank E Harrell Jr wrote:

....
> I'm sorry to have taken so long in responding to your excellent 
> question
> John.  And I'm responding to r-help since the question was posed there
> before taking the discussion offline.
>
> In the words of Don Berry "It takes time to be a Bayesian" and that's 
> the
> main reason there are no explicit Bayesian calculations in the book.  
> I do
> make a lot of use of prior information though.  In the future I could 
> see
> making some additions to the book along the lines of inclusion of 
> examples
> using the MCMCpack package, whose design is appealing to me.
>
>> R provides a lot of help for those who want a frequentist
>> interpretation, even to including by default the *, **, ***
>> labeling that some of us deplore.  There is no similar help
>> for those who want at least the opportunity to place the
>> output from a modeling exercise in a Bayesian context of
>> some description.  There is surely a strong argument for
>> the use of a more neutral form of default output, even to
>> the excluding of p-values, on the argument that they also
>> push too strongly in the direction of a frequentist
>> interpretative framework.
>
> Agreed.  I do try to approximate the Bayesian approach with the 
> bootstrap.
>
>>
>> There seems, unfortunately, to be a dearth of good ideas
>> on how the assist the placing of output from modeling
>> functions such as R provides in an explicitly Bayesian
>> framework.  . . . .
>
> It's all worth pursuing.  I wish there were already a Bayesian package
> that made use of Bayesian methods irresistable.
>
> All the best,
>
> Frank

John Maindonald             email: john.maindonald at anu.edu.au
phone : +61 2 (6125)3473    fax  : +61 2(6125)5549
Centre for Bioinformation Science, Room 1194,
John Dedman Mathematical Sciences Building (Building 27)
Australian National University, Canberra ACT 0200.