"anovan" seems to give the same number of degrees of freedom to a factor whether it is declared as random or fixed.
I thought that random factors "cost" less degrees of freedom. At least this is what happens in regression with mixed effects.
Any clues? Is this normal behavior?