On 15 Aug 2003 07:48:00 -0700, [EMAIL PROTECTED] wrote:

> I have a question concerning power and multiple regression. For my 
> study, I have 9 predictor variables, an alpha level of .05, and am 
> using .50 power. How would I estimate the number of participants for 
> my study? Thanks.
 
How to estimate the N?
I'd say:  At least, read all the last chapters of Cohen's book
on power analysis, and use his tables.

Since you are asking the question, I am pretty sure that you
need to read  *at least*  that much -- if you are to hope
to understand the answer.  

He's going to encourage you to figure out if you really
want nine predictors -- instead of one, or three.  And do
you really want to test them simultaneously? - do you
hypothesize patterns? - or would you have more power
(and much more intelligible tests)
if you set up univariate tests and Bonferroni correction?

-- 
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
"Taxes are the price we pay for civilization."  Justice Holmes.
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to