Consider the simultaneous analysis of a series of trials
with similar treatments and layout. For field trials,
the residuals are independent between trials but may be spatially
within trials. At the treatment level there is the
average effect of treatments across trials and the interaction
of treatment with trial.
For analysis in ASReml, the data for each trial needs to be
arranged conformably in a single or
multiple files.
The data also needs to be sorted into plot order
but this can be done by ASReml if the rows and columns
are coded in the data file. Assume we have a data field
site
which identifies the trial, and fields
row
and
column
which code for the spatial arrangement within trials.
Syntax
When
!SECTION site !ROWFAC row !COLFAC column
is specified, ASReml 4 checks that the nominated variables define a complete
grid of plots for each section. If any are missing, the data is augmented to 'complete the grid' for each section.
This is to facilitate fitting
a separable first-order autoregressive (
AR1 x AR1
) variance structure for each site.
column
is the name of a factor or variate containing column numbers (1 ...
nc where nc is the number of columns) on which the data is to
be sorted. row
is the name of a factor or variate containing row numbers (1 ...
nr where nr is the number of rows) on which the data is to
be sorted. site
is the
name of a factor or variate containing site numbers (1 ...
ns where ns is the number of sites) on which the data is already sorted.
The number of rows and columns within a trial does not
have to be the same across trials.
Previously the specification of these 3 qualifiers would generate
the R structure lines needed to fit a spatial model, but under the functional modelsyntax,
that functionality is not required because the spatial structure
is simply written as