On Fri, 13 Jul 2007, simone wrote:
> Hello everybody
>
> I have done a generalized linear model (glm-function) with a binary
> response variable (presence/absence) and two explanatory variables
> (including the interaction term (model<-glm(x~factor1*factor2,
> family=quasibinomial)). The interaction is what we are mainly
> interested in and it is non-significant (F=0, p=1.0). A reviewer on
> the manuscript now asks for a a priori power analysis on the non-
> significant interaction to see what sort of effect size we could
> reasonably have detected with our experimental design. As far as I
> have seen there is no power analysis function implemented for glm in
> R, or am I mistaken? Or does any one have any other suggestions how
> to deal with that comment?
Well, by citing
Hoenig, John M. and Heisey, Dennis M. (2001), ``The Abuse of Power: The
Pervasive Fallacy of Power Calculations for Data Analysis,'' The
American
Statistician, 55, 19-24.
and reviewing its argument for the benefit of the reviewer and the editor.
---
You have the data, you can compute the confidence interval (or region if
either factor has more than two levels), and then you will know just
what is ruled out by your non-significant interaction.
---
HTH,
Chuck
>
> I would really appreciate your help!
> Thanks in advance
>
> Simone
>
> Simone Haerri
> Institute of Environmental Sciences
> University of Zurich
> Winterthurerstrasse 190
> Ch - 8057 Zurich
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
Charles C. Berry (858) 534-2098
Dept of Family/Preventive Medicine
E mailto:cberry at tajo.ucsd.edu UC San Diego
http://famprevmed.ucsd.edu/faculty/cberry/ La Jolla, San Diego 92093-0901