Prediction following REML analysis in Genstat
by
Sue Welham
Biomathematics Unit, Rothamsted Research Ltd, Harpenden AL5 2JQ, UK
Coauthors: Arthur Gilmour (Orange Agricultural Institute, Orange, NSW 2800), Brian Cullis (Wagga Wagga Agricultural Institute, Wagga Wagga, NSW 2650), Bev Gogel (Queensland Department of Primary Industries, Yeerongpilly, Queensland 4299), Robin Thompson (Biomathematics Unit, Rothamsted Research Ltd, Harpenden AL5 2JQ, UK)

Linear mixed models are widely used in the analysis of biological data. Following estimation, it is often desirable to construct predicted values from the effects fitted in order to explore relationships established in the analysis. Gilmour et al (2002) describe an efficient algorithm for calculating predictions, or linear combinations of fixed and random effects, for a mixed model using REML analysis. This algorithm was recently implemented in ASREML (Gilmour et al, 1999) and has now been included within Genstat.

Prediction is already available in Genstat after fitting a regression or generalised linear model. However, some new concepts are required when forming predictions from a linear mixed model. This talk will briefly outline the additional considerations required when forming predictions in linear mixed models, and will describe the new directive, VPREDICT, that has been implemented to enable mixed model prediction in Genstat.

References

Gilmour AR, Cullis BR, Welham SJ, Gogel BJ, Thompson R (2002) An efficient strategy for prediction in mixed linear models. CSDA, to appear.

Gilmour AR, Cullis BR, Welham SJ, Thompson R (1999) ASREML Reference Manual. Biometric Bulletin 3, NSW Agriculture, Locked Bag 21, Orange, NSW 2800, Australia.

Date received: September 13, 2002

Copyright © 2002 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cajn-44.