Adds a covariance-structure definition to a REML model-definition structure (R.W. Payne).
Options
MODELSTRUCTURE = pointer |
Supplies the model-definition structure; no default (must be specified) |
|---|---|
EXPERIMENT = scalar |
Level of the EXPERIMENTS factor for which a residual is to be defined (using the VRESIDUAL directive) |
TERMS = formula |
Model terms for which the covariance structure is to be defined |
FORMATION = string token |
Whether the structure is formed by direct product construction or by definition of the whole matrix (direct, whole); default dire |
COORDINATES = identifiers |
Coordinates of the data points to be used in calculating distance-based models (list of variates or matrix) |
Parameters
MODELTYPE = string tokens |
Type of covariance model associated with the term(s), or with individual factors in the term(s) if FORMATION=direct (identity, fixed, AR, MA, ARMA, power, banded, correlation, antedependence, unstructured, diagonal, uniform, FA, FAequal) default iden |
|---|---|
ORDER = scalar |
Order of model |
HETEROGENEITY = string token |
Heterogeneity for correlation matrices (none, outside); default none |
METRIC = string token |
How to calculate distances when MODELTYPE=power (cityblock, squared, euclidean); default city |
FACTOR = factors |
Factors over which to form direct products |
Description
VFSTRUCTURE is one of a suite of procedures designed to simplify the assessment of alternative models for a REML analysis. The first step is to form a model-definition structure for each candidate model, using the VFMODEL and the VFSTRUCTURE procedures (these define the model settings controlled by the VCOMPONENTS and the VSTRUCTURE and VRESIDUAL directives, respectively). The model-definition structures can then be used as input to procedures like VARANDOM, which assesses possible random models. VARANDOM uses VMODEL to specify each model, in turn, so that it can fit it using REML. The relevant results from each fit are saved by the VRACCUMULATE procedure, so that a decision about the recommended random model can be made once they have all been tried.
The model-definition structure must be specified by the MODELDEFINITION option. Details of the model are specified by the TERMS, FORMATION, COORDINATES and EXPERIMENT options, and the MODELTYPE, ORDER, HETEROGENEITY, METRIC, and FACTOR parameters (which correspond to those options and parameters in the VSTRUCTURE and VRESIDUAL directives). If the EXPERIMENT option is not set, the specification will be used in a VSTRUCTURE statement within VMODEL. The EXPERIMENT option is relevant if you have used the EXPERIMENTS option in the original VFMODEL statement to define the experiments factor for a meta analysis. You can then set EXPERIMENT to a level of that factor to define a residual model for that experiment, using a VRESIDUAL statement within VMODEL.
Options: MODELSTRUCTURE, EXPERIMENT, TERMS, FORMATION, COORDINATES.
Parameters: MODELTYPE, ORDER, HETEROGENEITY, METRIC, FACTOR.
See also
Directives: REML, VCOMPONENTS, VSTRUCTURE.
Procedures: VARANDOM, VFMODEL, VMODEL.
Commands for: REML analysis of linear mixed models.
Example
CAPTION 'VFSTRUCTURE example',\
'Slate Hall Farm data (Guide to REML in Genstat, Section 1.8).';\
STYLE=meta,plain
SPLOAD '%gendir%/data/slatehall.gsh'
" define an Ar1 (x) Ar1 covariance model "
VFMODEL [MODELSTRUCTURE=AR1xAR1; DESCRIPTION='Ar1 (x) AR1';\
FIXED=variety] fieldrow.fieldcolumn
VFSTRUCTURE [MODELSTRUCTURE=AR1xAR1; TERMS=fieldrow.fieldcolumn]\
2('AR'); ORDER=1; FACTOR=fieldrow,fieldcolumn
VMODEL AR1xAR1
REML [PRINT=model,components,wald] yield