On Tue, 08 Nov 2011 06:25:24 -0800, Michael Britt wrote:
>This article looks timely:
>http://pss.sagepub.com/content/early/2011/10/17/0956797611417632.full 

The article that Michael B. links to is the following:

Joseph P. Simmons, Leif D. Nelson, and Uri Simonsohn
False-Positive Psychology: Undisclosed Flexibility in Data Collection and 
Analysis Allows Presenting Anything as Significant Psychological Science 
October 2011 0956797611417632, first published on October 17, 2011 
doi:10.1177/0956797611417632 
NOTE: A more recent version of this article was published on - Nov 7, 2011

Members of APS receive weekly listings of articles in the APS
journals and when the one for Psychological Science came this
morning, this was the first article I read (though I was tempted
to read the working memory article first).

The authors examine how researchers may engage in certain activities
that maximize the probability of getting a statistically significant result
but, it is argued, that this process actually produces more false positive
or Type I errors.  The authors use the phrase of "degrees of freedom"
in two different ways (1) the usual sense associated with a statistical
test, and (2) the number of ways that one can use variables, conduct
analyses, and other "methodological" procedures that can be used
to increase the likelihood of a statistically significant result and, by
extension, increase the Type I/false positive rate.

I am not totally convinced by the argument but I have only read the
article once.  Nonetheless, the authors propose 6 rules for researchers
or authors such as (quoting from the article)::

|1. Authors must decide the rule for terminating data
|collection before data collection begins and report
|this rule in the article.
|2. Authors must collect at least 20 observations per
|cell or else provide a compelling cost-of-datacollection
|justification.
|3. Authors must list all variables collected in a study
|4. Authors must report all experimental conditions,
|including failed manipulations.
|5. If observations are eliminated, authors must also
|report what the statistical results are if those
|observations are included.
|6. If an analysis includes a covariate, authors must
|report the statistical results of the analysis without
|the covariate.

There also are some suggested rules for journal reviewers but
I'll leave it to the interested reader to read them.

I think that a number of points made by Simmons et al are valid
but I think that their presentation is lacking.  I think that their
recommendations are somewhat reasonable but arbitrary (e.g.,
rule 2 above specifies 20 observations per cell but shouldn't
one conduct an a priori power analysis to determine what the
minimum sample size should be to detect a specific effect at
specific levels of power?).

I'd like to see what others think of the article.

-Mike Palij
New York University
[email protected]




---
You are currently subscribed to tips as: [email protected].
To unsubscribe click here: 
http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5&n=T&l=tips&o=13987
or send a blank email to 
leave-13987-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu

Reply via email to