高清pdf下载:
Introduction
In recent years a general algorithm, Restricted Maximum Likelihood (REML) has been
developed for estimating variance parameters in linear mixed models (LMM).
This manual will review classic statistical techniques (ANOVA & REGRESSION) and
demonstrate how LMM (REML) can be used to analyse normally distributed data from
virtually any situation. For balanced data, REML reproduces the statistics familiar to those
who use ANOVA, but the algorithm is not dependent on balance. It allows for spatial and/or
temporal correlations, so can be used for repeated measures or field-correlated data. Unlike
ANOVA, REML allows for changing variances, so can be used in experiments where some
treatments (for example different spacings, crops growing over time, treatments that include a
control) have a changing variance structure. The statistical package GenStat is used
throughout. The current version is 13, although the analyses can generally be performed
using the Discovery Edition released in 2010.