[R-lang] Re: anova formula repeated measures

[Roger Levy]

Hi Mike,

I have to admit that the proper specification of error terms for aov() has always been something of a thorn in my side.   Have you tried the following, though?

by subjects: aov(response ~ within1 * within2 * between + Error(between / (subject / (within1 * within2))))

by items: aov(response ~ within1 * within2 * between + Error(item / (within1 * within2 * between)))

(I should also note that I rarely do between subjects designs, and venture this suggestion with a low level of confidence...)

Best & good luck,

Roger

On May 20, 2010, at 5:33 PM, Michael T Hammond wrote:

> all
> 
> I am stuck on how to express a simple design for aov().
> 
> It's an experiment with two within-subjects factors (phonological
> properties of my stimuli) and one between-subjects factor (presentation
> mode: some subjects get things visually, some auditorily). There are
> multiple subjects of course, and multiple items in each factor and
> intersection.
> 
> If I left out the between-subjects factor, this is the normal way I would
> do this:
> 
> by items: aov(response ~ within1 * within2 + Error(item))
> 
> by subjects: aov(response ~ within1 * within2 + Error(subject/(within1 *
> within2)))
> 
> When I add in the between-subjects factor, I can't get the error term
> right:
> 
> by items: aov(response ~ within1 * within2 * between + Error(???))
> 
> by subjects: aov(response ~ within1 * within2 * between + Error(???))
> 
> I would have thought Error(item/between) and Error(subject/(within1 *
> within2)) respectively, but this does not work.
> 
> I've been searching the web and have not found any examples of this sort
> of mixed design with aov().
> 
> Does anybody have any ideas?
> 
> Mike H.

--

Roger Levy                      Email: rlevy@ling.ucsd.edu
Assistant Professor             Phone: 858-534-7219
Department of Linguistics       Fax:   858-534-4789
UC San Diego                    Web:   http://ling.ucsd.edu/~rlevy