Ok, I've taken another look at the paper (Lindberg et al, New 
Trends in Gender....Psych Bull, 2010).

I asked whether the average VR of 1.07 they reported in their 
first study was significantly different from a VR of 1.00 (and 
could have asked  this about their second study, where it was 
1.09).

This made me wonder how they were obtaining their average 
VR ratios. I looked at it for their second study, where they had 
four large data set studies, each reporting  VRs for multiple 
years.

Did they consider each VR a data point, and lump them all 
together in one big average? Or did they consider each study 
separately, and then average the averages? If the first, then 
eyeballing it suggests that their average VR of 1.09 would differ 
significantly from 1.00 over 56 separate VR data points. If they 
only used the average of each study (four of them), the number 
of entries would be too small for a meaningful answer.

But it seemed to me that lumping all the individual data points 
together was improper, and that they should only consider one 
a(average) VR value per study.

So which was it? What did they do? I tried to find out. For the 
one big average case (n= 56), I calculated the VR ratio from 
their data as 1.103. With the average of averages (n = 4) it 
came to 1.0975 . Neither is the 1.09 they reported but both are 
very close to it, and are very close to each other. [My 
understanding, together with a bit of algebra to make sure is 
that the overall average is not necessarily the same as the 
average of averages, but if I'm wrong, I'm going to look pretty 
silly here].

They did say they used the method of Katzman and Alliger 
(1992), whose title indicates it's a critique of methods of 
averaging variances, so perhaps they did neither of the above.

But the real news, which escaped me last time, is that these 
values of VR, however obtained, are not really that much lower 
from those she cites as earlier published estimates. So it's 
really, more or less (a bit less)  a replication of earlier claims of 
variability, not "nearly equal male and female variances" as she 
says in her abstract. As Jim Clark showed, this difference can 
mean a big deal at the extreme end of the tail. So it leaves 
unconvincing her conclusion that "these findings support the 
view that males and females perform similarly in mathematics"
.
Also, I'd like to amend this statement with which I ended my 
previous post:
> 
> For what it's worth, the hypothesis that seems most likely to me 
> is the self-selection one. Women may just not find full 
> professorship at Harvard in mathematics one of the most 
> fulfilling things they can do with their lives. That, of course, and 
> innate ability at the very, very high end. 
> .

I should add to that I can also readily believe that the good old 
boys at Harvard may well harbour a certain prejudice against 
hiring women. So, like Larry Summers, I hedge my bets.

Stephen

--------------------------------------------
Stephen L. Black, Ph.D.          
Professor of Psychology, Emeritus   
Bishop's University
Sherbrooke, Quebec, Canada               
e-mail:  sblack at ubishops.ca
---------------------------------------------

---
You are currently subscribed to tips as: arch...@jab.org.
To unsubscribe click here: 
http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5&n=T&l=tips&o=6274
or send a blank email to 
leave-6274-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu

Reply via email to