anova
,
which inherits from
data.frame
, or
miList
objects
whose components are of class
anova
.
miAnovaAux
is generic (see Methods);
method functions can be written to handle specific classes of data.
Classes which already have methods include:
(none yet).
miAnova(object, ...) miAnovaAux(object1, object, ...)
mi
object.
object
,
i.e.
miSubscript(object, 1)
.
miAnovaAux
,
and (depending on the method) ultimately to
anova
.
miList
whose components are
obtained by calling
anova
for each imputation.
miAnova
checks that all imputations have the same class,
then calls
miAnovaAux
, which dispatches based on that class.
The appropriate method to use in combining multiple imputations
usually depends not on the class of
object
but rather on the class
of the imputations of
object
. For example, suppose that
object
is
an
miList
object whose imputations are
lm
objects;
this generic function then dispatches to
miAnovaAux.lm
(if it exists),
otherwise to
miAnovaAux.default
.
Methods for
miAnovaAux
can be written to provide appropriate summaries
which consolidate information across multiple imputations.
Use
getMethods("miAnovaAux")
to see what methods have been defined
(currently none, because it is unclear how to properly
combine Anova tables across imputations).
If no method is found, then control passes to
miAnovaAux.default
,
which calls
anova
for each imputation, but does not consolidate
information across imputations.
Schafer, J. L. (1997), Analysis of Incomplete Multivariate Data , Chapman & Hall, London.
fit <- miEval(lm(chol14~., data = cholesterolImpExample)) miAnova(fit)