Page 1 of 1

empirical Bayes CIs for level-2 means

Posted: Thu Jul 21, 2011 4:52 am
by quecher
I'm using a random intercept null model for longitudinal data, with level 1 having 3-10 timepoints of measurements in individuals, which are level 2. I'm interested in predicting the 90% confidence intervals of each individual's mean. I'm familiar with using the Residuals and Predictions windows to calculate level-2 means and residuals, but I'd like the "empirical Bayes" predictions of the confidence intervals.

When I estimated these in SAS Proc Mixed, the confidence intervals for individual means were not always evenly spread above and below the mean, so they aren't equivalent to simply multiplying each standard error of the level-2 predicted means by 1.96. I've tried using the Customised Predictions window to no avail. Is there a way to get these types of results without using MCMC?