Teach decision theory
-------- Original Message --------
Subject: [SPAM] Re: BLOG: Guest post - Don't teach hypothesis testing
From: Nicola Petty <[log in to unmask]>
Date: Mon, February 02, 2015 8:35 am
To: [log in to unmask]
Hi all
Sorry to email again so soon.
The point of the post mentioned in the title here is for discussion, and
I’m sure Tony will be pleased to hear from you. I don’t think he is
on the list, so I have forwarded any comments to him.
It would be better for the discussion to occur on the blog, where other
people can benefit from it, and we don’t annoy people who don’t want
discussion on the allstat list.
Best wishes
Nicola
From: A UK-based worldwide e-mail broadcast system mailing list
[mailto:[log in to unmask]] On Behalf Of Martin Holt
Sent: Tuesday, 3 February 2015 3:00 a.m.
To: [log in to unmask]
Subject: Re: BLOG: Guest post - Don't teach hypothesis testing
Dear Nicolas
I know that allstat does not welcome publishing discussions on topics
such as this. A while ago I founded the Google Group MedStats with the
intention of full disclosure of all the discussion (anyone can view the
discussion; you need to join to post a query.) The group now numbers
over 1300 and hosts some of the most notable statisticians: Bland,
Altman, Frank Harrell (Jnr), Campbell to name but a few.
Especially with your posting being so interesting, I would
wholeheartedly recommend that you join MedStats and post it there.
Kind Regards,
Martin
Martin P. Holt
Freelance Medical Statistician and Quality Expert
[log in to unmask]
Persistence and Determination Alone are Omnipotent !
If you can't explain it simply, you don't understand it well
enough.....Einstein
Linked In:
https://www.linkedin.com/profile/edit?trk=nav_responsive_sub_nav_edit_profile
From: Nicolas <[log in to unmask]>
To: [log in to unmask]
Sent: Monday, 2 February 2015, 13:09
Subject: Re: BLOG: Guest post - Don't teach hypothesis testing
Hi everyone,
I agree that "null hypothesis significance testing" needs to be
rethought (dropped or at least taught differently).
However I do think the p-value is still a very useful measure, because
it adds the information of the magnitude of our surprise under H0. This
information cannot be found in a confidence interval.
For example, if the density distribution of our estimate under the null
hypothesis is extremely low in 0 or below (say p-value in 0 is
0.00000001), increases very rapidly between 0 and 1 and then slowly
decreases when greater than 1, we could have a 95% confidence interval
of, say [0.1, 27.5], where one may think that 0 is "close" to be in our
confidence interval, whereas it is actually very far from being inside.
Of course one could then provide a 99.9999% confidence interval that
would not include 0, but I feel such confidence interval would more
easily confuse the reader than the p-value.
Don't you think so?
Cheers,
Nicolas
Date: Mon, 2 Feb 2015 09:02:05 +0000
From: [log in to unmask]
Subject: Re: BLOG: Guest post - Don't teach hypothesis testing
To: [log in to unmask]
It all depends whom you are teaching, of course. However, in an
introductory course I wd say
1 start with EDA and then go to confidence intervals using the median
2 thence to hypothesis testing.
3 Give a small number of robust procedures, rather than lots of fiddly
baffling ones.
4 Don't forget that statistics has changed since we leArned the subject!
JOHN. BIBBY
On Monday, 2 February 2015, Nicola Petty <[log in to unmask]> wrote:
Hi all
This post is from Tony Hak on a topic of concern to many of us. Are we
right to still keep teaching null hypothesis significance testing?
Tony gives three arguments against it, and would really like to hear
from the community.
http://wp.me/p24HeL-mc
Enjoy
Nicola
===========================================================
Dr Nicola Ward Petty
Aka Dr Nic
Mobile: 027 698 7392
Landline: 03 980 0901
Statistics Learning Centre
http://statslc.com/
Learning and teaching statistics is surprisingly interesting with Dr
Nic.
You may leave the list at any time by sending the command SIGNOFF
allstat
to [log in to unmask], leaving the subject line blank.
You may leave the list at any time by sending the command SIGNOFF
allstat
to [log in to unmask], leaving the subject line blank.
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
You may leave the list at any time by sending the command SIGNOFF
allstat
to [log in to unmask], leaving the subject line blank.
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|
