to estimate missing values in the response variable. Formally this
creates a model term with a column for each missing value. Each column
contains zeros except for a solitary -1 in the record containing the
corresponding missing value.
The equations for mv and any
terms that follow are always included in the sparse set of equations.
are handled in three possible
ways during analysis. In the simplest case, records containing missing
values in the response variable are deleted. For multivariate
(including some repeated measures) analysis, records with missing values
are not deleted but ASReml drops the missing observation and uses
the appropriate unstructured R-inverse matrix. For regular spatial analysis,
we prefer to retain separability and therefore estimate the missing value(s)
by including the special term
mv in the model.
Return to start