[R] ANOVA non-sphericity test and corrections (eg, Greenhouse-Geisser)
s.blomberg1 at uq.edu.au
Mon Jun 25 09:53:59 CEST 2007
If you use lme, you can fit a general correlation structure to the
within-subject data, and compare the fit to a model assuming
uncorrelated within-subjects errors. That should tell you whether your
data are Aren't the G-G and H-F corrections only approximate fixes?
Surely it is better to work with a model that actually fits your data,
rather than using ad hoc adjustments towards a model that doesn't quite
fit. But I'm no psychologist. :-)
On Mon, 2007-06-25 at 08:22 +0200, Peter Dalgaard wrote:
> DarrenWeber wrote:
> > I'm an experimental psychologist and when I run ANOVA analysis in
> > SPSS, I normally ask for a test of non-sphericity (Box's M-test). I
> > also ask for output of the corrections for non-sphericity, such as
> > Greenhouse-Geisser and Huhn-Feldt. These tests and correction factors
> > are commonly used in the journals for experimental and other
> > psychology reports. I have been switching from SPSS to R for over a
> > year now, but I realize now that I don't have the non-sphericity test
> > and correction factors.
> This can be done using anova.mlm() and mauchly.test() which work on
> "mlm" objects, i.e., lm() output where the response is a matrix. There
> is no theory, to my knowledge, to support it for general aov() models,
> the catch being that you need to have a within-subject covariance matrix.
> R-help at stat.math.ethz.ch mailing list
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
Simon Blomberg, BSc (Hons), PhD, MAppStat.
Lecturer and Consultant Statistician
Faculty of Biological and Chemical Sciences
The University of Queensland
St. Lucia Queensland 4072
Room 320, Goddard Building (8)
T: +61 7 3365 2506
The combination of some data and an aching desire for
an answer does not ensure that a reasonable answer can
be extracted from a given body of data. - John Tukey.
More information about the R-help