On 2012-02-22, at 8:57 AM, Marc Carter wrote:

> In ANOVA, why are the squared deviations in the SS between groups scaled 
> (multiplied) by _n_?  I was once told it was to weight them, but that somehow 
> doesn't seem right.
> 

Marc,

Essentially, you are estimating the error variance in two different ways: (1) 
finding the mean of the group variances, (2) finding the variance of the group 
means. BUT, the variance of the group means doesn't estimate the error variance 
of the data, you'll notice. It estimates the (square of the) standard error of 
the mean (SEM). In order to convert an SEM^2 to a data variance, you have to 
multiply it by n (just look at the formula for SEM and figure out how to 
extract the variance in the numerator: multiply it by n to cancel out the n in 
the denominator).

After you have estimated the error variance in these two ways, you will notice 
that the 2nd way is sensitive to differences between the means, but the first 
way is not sensitive to such differences. So, if you divide the second estimate 
by the first, and the number is substantially higher than 1.0, you have reason 
to believe that the difference was generated by differences among the group 
means (presuming, of course, that you have satisfied all the various 
"assumptions" of ANOVA, which are really just ways of eliminating alternative 
explanations for any difference found  between the two estimates).

David Howell's text explains this really well. 


Best,
Chris
---
Christopher D. Green
Department of Psychology
York University
Toronto, ON M3J 1P3
Canada

[email protected]
http://www.yorku.ca/christo/
==========================
---
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=16208
or send a blank email to 
leave-16208-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu

Reply via email to