Skip to content
Prev 677 / 20628 Next

Split-plot Design

A response to your final paragraph.  There's no reason in principle
why aov() should not allow for random slopes.  In fact, one can
put continuous variables into the Error() term, but without taking
time to study the output with some care, it is not obvious to me what
the results mean.

[For reasons that are not clear to me, you mentioned Type III
"errors".  The controversy over Type III versus Type I sums of
squares relates to models with fixed effect interactions.

With models that have a single error term, drop1() will example
the effect of single term deletions, often in unbalanced designs
more relevant than the sequential information in the anova table.
Importantly, drop1() respects hierarchy, whereas Type III sums of
squares do not.]

John Maindonald             email: john.maindonald at anu.edu.au
phone : +61 2 (6125)3473    fax  : +61 2(6125)5549
Centre for Mathematics & Its Applications, Room 1194,
John Dedman Mathematical Sciences Building (Building 27)
Australian National University, Canberra ACT 0200.
On 22 Mar 2008, at 12:51 AM, Andy Fugard wrote: