Dear all,
I'm having trouble getting 95% confidence intervals for the estimates of fixed effects with using lmer when doing some polynomial regression (3rd order max).
mod1 being my model fitted following (too many data points to show):
mod1 <- lmer(y~poly(x,3)+(1|group)+(poly(x,3)-1|group))
running mcmcsamp(mod1,5000) returns:
Error in .local(object, n, verbose, ...) :
Code for non-trivial theta_T not yet written
(this is independent of the raw argument in poly())
I understand that this means that the code for dealing with correlated random effect is not written yet (https://stat.ethz.ch/pipermail/r-sig-mixed-models/2008q3/001410.html).
I haven't found any suggestions other than the implicit 'wait for it to be written'.
Should I forget about stochastic determination of the CI and use t-values, guessing the degrees of freedom?
Any hint would be appreciated, and generally thanks for the support that the list archives provides.
simon
PS/ I don't speak winbugs, but this could be an opportunity if the solution lies there.
[[alternative HTML version deleted]]