Re: st: power repeated measures anova vs mixed models

[Ricardo Ovaldia <ovaldia@yahoo.com>]

Thank you Austin for the insightful remark: "Of course what you plug in matters"

You are missing the point. I have the sample size (n=65/group), power (80%) and alpha (5%), 3 groups and 6 time points. What I want to compute is the minimal detectable effect size. I did the power analysis using a repeated measure ANOVA and obtained the minimal detectable effect sizes assuming various correlations between the repeated measurements. What I want to know is whether the mixed model would have more power to detect these effect sizes? 

Ricardo Ovaldia, MS
Statistician 
Oklahoma City, OK


--- On Thu, 5/24/12, Austin Nichols <austinnichols@gmail.com> wrote:

> From: Austin Nichols <austinnichols@gmail.com>
> Subject: Re: st: power repeated measures anova vs mixed models
> To: statalist@hsphsun2.harvard.edu
> Date: Thursday, May 24, 2012, 11:59 AM
> Ricardo Ovaldia <ovaldia@yahoo.com>:
> Of course what you plug in matters--see also:
> http://www.urban.org/UploadedPDF/1001394-clustered-randomization.pdf
> (esp. page 6) and references therein.
> You need estimates for the relevant information before you
> can estimate power.
> 
> On Thu, May 24, 2012 at 12:39 PM, Ricardo Ovaldia <ovaldia@yahoo.com>
> wrote:
> > Thank you David.
> > I played with this a few day ago. The problem is that
> you have to make a lot of assumptions that I do not feel
> comfortably making because I lack prior knowledge about
> parameters, covariances, etc. The program produces very
> different results depending on what you "plug in".
> >
> > Ricardo
> >
> > Ricardo Ovaldia, MS
> > Statistician
> > Oklahoma City, OK
> >
> >
> > --- On Thu, 5/24/12, Airey, David C <david.airey@vanderbilt.edu>
> wrote:
> >
> >> From: Airey, David C <david.airey@vanderbilt.edu>
> >> Subject: re: st: power repeated measures anova vs
> mixed models
> >> To: "statalist@hsphsun2.harvard.edu"
> <statalist@hsphsun2.harvard.edu>
> >> Date: Thursday, May 24, 2012, 10:01 AM
> >> .
> >>
> >> I just came across this software for longitudinal
> /
> >> hierarchical experimental design power analysis:
> >>
> >> http://sitemaker.umich.edu/group-based/optimal_design_software
> >>
> >> I've not used it, but it might help you avoid
> simulation.
> >>
> >> -Dave
> >>
> >> > Dear all,
> >> >
> >> > I have been struggling to find an answer or
> reference
> >> to this problem.
> >> >
> >> > I am planning a longitudinal analysis
> comparing 3
> >> groups with 6 time points per subject. The design
> is balance
> >> with 65 subjects for group.
> >> >
> >> > Because I do not have preliminary data and do
> not want
> >> to make unrealistic assumptions about the
> covariate
> >> structure and other parameters required to
> calculate power
> >> for mixed models, I decided to use repeated
> measures ANOVA
> >> to estimate the minimum detectable effect size at
> 80% power.
> >>
> >> > My questions are, will the mixed model have
> more power
> >> that the repeated measures ANOVA in this case? Are
> there any
> >> references regarding these comparisons?
> >> >
> >> > Thank you,
> >> > Ricardo
> >> >
> >> > Ricardo Ovaldia, MS
> >> > Statistician
> >> > Oklahoma City, OK
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
> 

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/