I am trying to understand how the principles of nested anova in R, but I am having extreme difficulty with the error term.

Imagining the simplest case. I measured plant growth in two places, A and B. Inside place A, I applied treatment w to five plants, and treatment x to another five. In place B, I applied treatment y to five plants, and z to another five. In my understanding treatment is nested within place. I want to test the hypothesis that both place and treatment have significant effects, if possible.

I have seen many formulas so far, but never a explanation of the logic behind it:

1) aov(growth ~ treatment + Error(place) )
2) aov(growth ~ treatment + Error(place/treatment) )
3) aov(growth ~ treatment + Error(place:treatment) )
4) aov(growth ~ treatment * place + Error(place/treatment) )

What is the difference between these formulas?

I understand the meaning of a fixed interaction "*". I also saw many many answers recommending using linear mixed models, so I understand this is an archaic method.

However, I have no idea of what the terms inside Error() means. I can't even understand how many stratums should I expect, or which variables will have p-values.

Adding a third variable also increases my nightmare. It seems that some variables end in on the top stratum, some on the Within an others in the all "stratums". How can I understand this variable placement and how it is determined by the formula?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.