Newton-Raphson and EM algorithms for linear mixed-effects models for repeated-measures data. (English) Zbl 0671.65119

The authors consider linear mixed-effect models for repeated-measures data. They are concentrated mainly on the effective implementation of the Newton-Raphson (NR) algorithm for estimating the parameters and compare him with the EM-algorithm proposed for this model by N. H. Laird and J. H. Ware [Biometrics 38, 963-974 (1982; Zbl 0512.62107)]. They conclude that in most situations the well implemented NR-algorithm is preferable to the EM-algorithm. The authors propose also several new methods for an implementation both of the EM- and NR-algorithms, draw conclusions about their performance and discuss extensions of the mixed- effect model to incorporate nonindependent conditional error structure and nested-type designs.
Reviewer: J.Antoch


65C99 Probabilistic methods, stochastic differential equations
62J99 Linear inference, regression


Zbl 0512.62107
Full Text: DOI