I have a data frame with this structure:
'data.frame': 39 obs. of 3 variables:
$ topic : Factor w/ 13 levels "Acido Folico",..: 1 2 3 4 5 6 7 8 9 10 ...
$ variable: Factor w/ 3 levels "Both","Preconception",..: 1 1 1 1 1 1 1 1 1 1 ...
$ value : int 14 1 36 17 5 9 19 9 19 25 ...
and i want to test the effect value ~ variable, considering that observation are grouped in topics. So I thought to use a repeated measure ANOVA, where "variable" is considered as a repeatead measure on every topic. the call is aov(value ~ variable + Error(topic/variable)).
My question is: if i want to do the same with lme how should I model the random effect?
~ 1 | topic/variable
or
~ variable | topic
the anova returns a significant F. for the lme, the first version gives a non significant effect, while the second is significant; so I suppose that the anova is more similar to the second.
Did I understood right?
Thanks