APS: New Statistics

[Salbod]

APS made available this link on 'New Statistics'.

http://www.psychologicalscience.org/index.php/members/new-statistics

It is to prompt a book, but the videos are well done and informative.

Stephen Salbod, Pace University, NYC

[David Marso]

I don't see how any of this is "new".
I was studying meta-analysis in the late 80's?
Effect sizes and CI's long before that.

Salbod wrote

APS made available this link on 'New Statistics'.

http://www.psychologicalscience.org/index.php/members/new-statistics

It is to prompt a book, but the videos are well done and informative.

Stephen Salbod, Pace University, NYC

[Mike]

A few points:

(1)  Geoff Cumming who is featured in the videos and is
author of the book "Understanding the New Statistics" is part
of the movement in psychology to leave null hypothesis testing
in the dustbin of history and put in its place the "new" statistics
of effect sizes, confidence intervals, and meta-analysis.  He
has written a nice book on these topics and provides access to
a neat little program that allows one to fool around with these
techniques.  His book is available on Amazon and you can
see the positive endorsements made by a number of people,
including some readily recognizable statisticians; see:
http://www.amazon.com/Understanding-New-Statistics-Meta-Analysis-Multivariate/dp/041587968X/ref=sr_1_1?ie=UTF8&qid=1415741837&sr=8-1&keywords=geoff+cumming

(2) However, as David Marso has pointed out, there is really
very little new in what Cumming presents and, as I point out
in my review of his book, there are potential pitfalls that his
approach are subject to (e.g., the "decline effect", that is,
a statistically significant result that become less significant
over time or, equivalently, an effect size that goes to zero
with replications).  Since I have pimped Cumming's book,
allow me to do the same for my review which was published
in PsycCritiques:
https://www.researchgate.net/publication/236866116_New_statistical_rituals_for_old

(3) Please, no religious wars over whether null hypothesis testing
is justified or not (Cumming thinks it rots one's mind, that is,
leads to bad reasoning).  See my quote of mathematical psychologist
David Krantz who wrote a review of the book "What if There
were No Significance Tests?" for the Journal of the American
Statistical Association.  There are additional points about the
"tastes great" vs "less filling" debate (from the old Miller Lite beer
ads)
concerning whether one should use null hypothesis testing or
focus on confidence intervals/effects size/etc -- a real statistician
would ask which technique is most informative for the question
being asked.  One would then use null hypothesis testing and/or
confidence intervals/effect size/etc and/or Bayesian techniques
and/or model the phenomenon and work out the appropriate
tests yourself and so on.  In the end, there may be no simple answers
and no substitute for hard work and deep thinking.  See the
recent work on the problem of not being able to replicate results
(significant or not).

-Mike Palij
New York University
[hidden email]




----- Original Message -----
From: "Salbod" <[hidden email]>
To: <[hidden email]>
Sent: Tuesday, November 11, 2014 10:37 AM
Subject: APS: New Statistics


> APS made available this link on 'New Statistics'.
>
> http://www.psychologicalscience.org/index.php/members/new-statistics
>
> It is to prompt a book, but the videos are well done and informative.
>
> Stephen Salbod, Pace University, NYC

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bruce Weaver]

That's a nice review, Mike.  Thanks for sharing.

Mike Palij wrote

A few points:

(1)  Geoff Cumming who is featured in the videos and is
author of the book "Understanding the New Statistics" is part
of the movement in psychology to leave null hypothesis testing
in the dustbin of history and put in its place the "new" statistics
of effect sizes, confidence intervals, and meta-analysis.  He
has written a nice book on these topics and provides access to
a neat little program that allows one to fool around with these
techniques.  His book is available on Amazon and you can
see the positive endorsements made by a number of people,
including some readily recognizable statisticians; see:
http://www.amazon.com/Understanding-New-Statistics-Meta-Analysis-Multivariate/dp/041587968X/ref=sr_1_1?ie=UTF8&qid=1415741837&sr=8-1&keywords=geoff+cumming

(2) However, as David Marso has pointed out, there is really
very little new in what Cumming presents and, as I point out
in my review of his book, there are potential pitfalls that his
approach are subject to (e.g., the "decline effect", that is,
a statistically significant result that become less significant
over time or, equivalently, an effect size that goes to zero
with replications).  Since I have pimped Cumming's book,
allow me to do the same for my review which was published
in PsycCritiques:
https://www.researchgate.net/publication/236866116_New_statistical_rituals_for_old

(3) Please, no religious wars over whether null hypothesis testing
is justified or not (Cumming thinks it rots one's mind, that is,
leads to bad reasoning).  See my quote of mathematical psychologist
David Krantz who wrote a review of the book "What if There
were No Significance Tests?" for the Journal of the American
Statistical Association.  There are additional points about the
"tastes great" vs "less filling" debate (from the old Miller Lite beer
ads)
concerning whether one should use null hypothesis testing or
focus on confidence intervals/effects size/etc -- a real statistician
would ask which technique is most informative for the question
being asked.  One would then use null hypothesis testing and/or
confidence intervals/effect size/etc and/or Bayesian techniques
and/or model the phenomenon and work out the appropriate
tests yourself and so on.  In the end, there may be no simple answers
and no substitute for hard work and deep thinking.  See the
recent work on the problem of not being able to replicate results
(significant or not).

-Mike Palij
New York University
[hidden email]




----- Original Message -----
From: "Salbod" <[hidden email]>
To: <[hidden email]>
Sent: Tuesday, November 11, 2014 10:37 AM
Subject: APS: New Statistics


> APS made available this link on 'New Statistics'.
>
> http://www.psychologicalscience.org/index.php/members/new-statistics
>
> It is to prompt a book, but the videos are well done and informative.
>
> Stephen Salbod, Pace University, NYC

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD