Thanks to all who contributed to the answer to the stats question. I ended up 
doing the analyses both with the McNemar and with a paired samples t-test (time 
1 versus time 2) and then did a Simes' correction (a type of Bonferroni 
correction that I like better than most others because it is slightly more 
conservative so that I don't over-invest in my data) for both. It turned out 
almost exactly the same. Only one item differed and it goes back to the more 
conservative nature of the Simes' correction. This was MOST helpful to me as 
was the discussion of how to think about 1/0 data sets depending on the basis 
on which we assign the the 1s and 0s. I think I was over-thinking; my initial 
approach was the t-test approach with correction. But then I got to thinking 
that 1/0 data should be nominal in the sense of what is correct and what is 
incorrect is nominal in a sense; on the other hand, there are distinct criteria 
for determining accuracy that are clearly different from something, like, say 
gender which is almost immutable in its differentiation.

Thanks again for the great overall discussion of the finer points of this 
issue. I really love tips :) Thanks, Bill!

Annette

Annette Kujawski Taylor, Ph. D.
Professor, Psychological Sciences
University of San Diego
5998 Alcala Park
San Diego, CA 92110
[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=23206
or send a blank email to 
leave-23206-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu

Reply via email to