Exploration & description
Contents
Fixed and random effects
Observational units
ANOVA or REML?
Before incorporating ram and
ewe random effects into the statistical model it is worth
discussing first the meaning of mixed models.
Mixed model
methodology takes its name from the understanding that the
elements of the model underlying a statistical analysis can be
a mixture of what are called fixed and random effects. The
approach has become important in the analysis of data that
have a hierarchical structure, since the different layers in
the structure can be modelled using random effects.
A fundamental step in using mixed models for hierarchical data
is to recognise the structure, namely the different layers in
the data. In order to help with this we shall use what we
describe as a ´mixed model tree´ to develop the different
layers pictorially. This is also illustrated in the
statistical guide by Allan and Rowlands (2001) which uses the
data from this case study for one of its examples.
This guide is no.19 of the Good Practice Guides. It also includes examples from Case Study 6 and from the paper by Methu et al (2001).
|