嵌套在R中的3种方差分析

Question

I'm trying to figure out the model for a fully factorial experiment.

I have the following factors

Treatment   Day   Hour  Subject   ResponseVariable

10 days of measurements, 4 different time points within each day, 2 different treatments measured, 12 subjects )6 subjects within treatment 1, and 6 different subjects in treatment 2)

for each day I measured: 6 subjects in treatment 1, the other 6 in treatment 2, at 4 different time points.
For Subjects, I have 12 different subjects, but Subjects 1-6 are in Treatment-1 and Subjects 7-12 are in Treatment-2. The subjects did not change treatments, thus I measured the same set of subjects for each treatment each of the 10 days

So what's tripping me up is specifying the correct error term.
I thought I had the general model down but R is giving me "Error() model is singular"

aov(ResponseVariable  ~ T + R + S + TR + TS + RS + Error(T/S)

any thoughts would help?

Answer 1

I've gotten the same error, and I think my problem was missing observations. Are you missing any observations? I believe they're less of a problem for linear mixed effects, and I've read that some people use lme instead of repeated-measures ANOVA for those cases.

Your error term can be interpreted as "the S effect within each T". It sounds from your description as though that's what you want, so I don't think that's what's causing your error message.

One note: I see you've got a variable named "T". R let you do that? T is normally reserved for meaning "TRUE". That might be part of your problem.