Re: gpower:"other f tests"-Option: Which n and df?
- From: "Thom" <tsbaguley@xxxxxxxxx>
- Date: 19 Jul 2006 04:42:43 -0700
Nina H. wrote:
I am trying to calculate the power (1-beta) of a multivariate repeated
measures ANOVA with the programm "gpower".
Can anyone tell me which "n" I have to use for sample size? If 40
people are each tested twice, is n=40, the actual sample size, or n=80
because of the repeated measure? And if tested three times, would that
be n=120??
I also don't know which df-value I have to use as "numerator df" and
"denominator df". Would this be the effect- (p-1) and error (n-1)
degrees of freedom?
(Unfortunately, I do not understand the manual's instructions on this
issue.)
I don't think GPower does multivaraite analyses - also I'm not sure
Gpower is reallly useful for repeated measures ANOVA as the true power
depends on the covariance matrix amongst other things.
I briefly mention this in my piece on sphericity - which also cites and
links to Jeremy Miles's paper on power that explains how to use SEM
software to estimate power for MANOVA and repeated measures designs
(among other things):
http://psychologicalstatistics.blogspot.com/2006/05/what-is-all-this-stuff-about.html
A better approach is probably to specify your main hypotheses a priori
as 1 d.f. contrasts. this is a more powerful approach than multiple
d.f. ANOVA tests and in the repeated measures case means that they
reduce to paired or one-sample t tests. This means that power can be
readily estimated from pilot data (though small sample pilots tend to
overestimate power) or by using reasonable values ofthe difference in
means, SD and correlation between samples.
Thom
.
- Follow-Ups:
- Re: gpower: Does post hoc power analysis make sense?
- From: Nina H.
- Re: gpower: Does post hoc power analysis make sense?
- References:
- gpower:"other f tests"-Option: Which n and df?
- From: Nina H.
- gpower:"other f tests"-Option: Which n and df?
- Prev by Date: gpower:"other f tests"-Option: Which n and df?
- Next by Date: Re: Chi Square
- Previous by thread: gpower:"other f tests"-Option: Which n and df?
- Next by thread: Re: gpower: Does post hoc power analysis make sense?
- Index(es):
Relevant Pages
|
|
