Hello r-help,
I am using lme with two specs for the variance func
varComb(varFixed(~1/n)),varPower(~Age))
this produces worse forecasts than the lm model with simple
weights=n
I think due to the fact that the lme spec works on variance inside the group. I
need to show it that 1/n scales the variance between groups.
Is that possible?
I cannot disclose my dataset, but could post plots if that is possible somehow,
let me know. Anova shows that everything in random=~poly(age,2) is significant.
Thank you all very much.
Stephen
[[alternative HTML version deleted]]