very interesting.
the authors said
Often, exact p values were not reported
and were therefore calculated (e.g., using F values
and degrees of freedom). All p values within the
range of interest were recorded.
I wonder if the results would have varied if the authors had
used a slightly different approach.
1) used 3 searchers consistently and/or used computer aided
text processing to locate the p or test stats. (Were tests
reported without p's?)
2) recorded test values(F, t, r, etc.) and df a numbers as well
as the original reported values as a string.
3) then used a single package's p routines so that a) p finding
algorithm, b) rounding, and c) interval coarsening were done
consistently. e.g., CDF.*
4) spot eye-balled the string p to the recalculated p.
5) it might also be interesting to systematically look at the
printed p and the recalculated p.
wrt #3 it is astonishing how often people find their p-value by
eyeballing a table.
I wonder how people on this list would
apply what is known in psychology as the "so what test" to the
size of the discrepancy in p. Something analogous to the
"so what test" we apply to effect sizes.
P.S. If I had been a reviewer for this as a submitted article I
would have suggested that the editor reject until revised at least
up to step 3.
Art Kendall
Social Research Consultants
On 8/18/2012 4:21 AM, Albert-Jan Roskam wrote:
Hi all,
Thought I'd share this with you. Check out Figure 2. This
is shocking, although, sadly, not suprising at all:
E. J. Masicampo & Daniel R. Lalande (2012). A peculiar
prevalence of p values just below .05. The Quarterly Journal
of Experimental Psychology, 2 aug 2012
Regards,
Albert-Jan
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
All right, but apart from the sanitation, the medicine,
education, wine, public order, irrigation, roads, a
fresh water system, and public health, what have the Romans
ever done for us?
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
=====================
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
Art Kendall
Social Research Consultants
Locked
