Prediction Intervals For The Generalized Linear Mixed Model'

• Activity Number: 302
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Computing
• Author(s): Chenghsueh Yang*+
• Companies: UCR
• Keywords: best linear predictor ; best linear unbiased predictor ; generalized linear mixed model ; pseudo likelihood ; prediction interval ; quadrature

Abstract

Some illustrative generalized linear models are used to evaluate the performance (in terms of the coverage probability and the expected width) of alternative prediction intervals for the random effects. A prediction interval based on the eBLUP (empirical Best Linear Unbiased Predictor) that results from the Pseudo-Likelihood (PL) method is a commonly used interval. However, convergence of PL is sometimes not achieved. Alternatively, use of quadrature rules to fit the model is sometimes more viable and does not require as many assumptions as the PL method. The parameter estimates based on the quadrature method can be used to create a prediction interval based on Bayes or best linear predictors. There is an appreciable computational advantage associated with using best linear predictors.