[R-sig-ME] Variances and standard errors of predicted values following lmer()

[matthias.suter at art.admin.ch]

Dear R users

Does anybody know how to receive the variances (and standard errors) of predicted values following a fit with lmer()? I should only consider the variance of predicted values corresponding to the fixed estimates (I need no prediction to the random part). To my knowledge, a predict() function to an lmer() object is not implemented yet?

Given a matrix of new predictors (matrix of new data corresponding to the fixed effects), and the variance-covariance matrices of fixed (vcov) and random (VarCorr) effects from the lmer.output, it should be possible the calculate the variances of predicted values. Can somebody help?

Matthias Suter,
Agroscope Reckenholz-Tänikon ART, Zürich