[cc'ing to r-sig-mixed]
On 13-08-02 09:19 AM, michelle morters wrote:
Hi - Please may I ask how to fit CIs to a predicted value from my model. I've given up on lmer and have tried lme but still no luck. Below is a snippet of my data - it is very unbalanced. Time is in days and id = individual. The model with the lowest AIC is log(titre)~time+I(time^2),random=~1|id The geometric mean titre from my data at day 330 is 0.6IU When I manually plug day 330 into the model equation I get a GMT of 0.68 IU (great!) model: y=2.592-3.03e-2(330)+6.44e-5(330^2) However I need to report CIs for the predicted values. I'm stuck with both the predict function and the correct argument in the function for CIs - I have spent ages looking at manuals and trying to extrapolate worked on-line examples to my data but alas to no avail although I'm sure the solution is very simple! - please could I have some advice
There is no way to get confidence intervals for predictions from predict() for either lme or lmer models. At present the best solution (which is still very limited) is to use the recipe on http://glmm.wikidot.com/faq#predconf . If you get stuck, please post to r-sig-mixed-models explaining what you tried and how far you got.
Very many thanks, Michelle id time titre 1 0 2.83 1 60 0.5 1 150 0.25 2 0 2.83 2 60 0.09 2 150 0.25 2 330 0.25 3 0 5.66 3 60 0.18 5 0 22.63 5 60 1 5 150 1.41 5 330 4 6 0 16 6 60 0.71 6 150 0.5 6 330 0.25 7 0 90.51 8 0 11.31 8 60 2 8 150 1 9 0 22.63 9 60 2 9 150 0.5 9 330 0.71 10 0 0.09 10 60 0.25 10 150 0.18 10 330 0.06
_______________________________________________ Lme4-authors mailing list Lme4-authors at lists.r-forge.r-project.org https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/lme4-authors