miVariable
object.
miVariable(x, Imputations, whichNA = which.na(x), subset = F, Names = 1:length(Imputations), row.names) as.miVariable(x)
miVariable
): atomic (non-list) variable.
as.miVariable
):
miList
object whose
imputations are atomic variables with the same length, mode, and attributes,
or an
miVariable
object.
More general
mi
objects are allowed, but the result will
not be an
miVariable
object.
x
(or equal to the length of
x
, if
subset=TRUE
).
x
are
missing.
TRUE
, then each element of
Imputations
will be replaced with
the subset indicated by
whichNA
.
Imputations
data frame in the returned value. These are used if supplied explicitly,
otherwise if the
Imputations
argument has names they are used,
otherwise default names are used.
Imputations
data frame in the returned value.
"miVariable"
, with slots:
x
to
miVariable
; this may have its
own class),
whichNA
and as many columns
as there are sets of multiple imputations.
as.miVariable
, if
x
cannot
be converted to an
miVariable
object, then a more general
mi
object will be returned.
The
whichNA
argument is typically equal to
which.na(x)
or
is.na(x)
.
x
may be a matrix or array.
Schafer, J. L. (1997), Analysis of Incomplete Multivariate Data , Chapman & Hall, London.
c3 <- miEval(crimeImpExample[[3]]) c3 # an miList object, 10 sets of imputations as.miVariable(c3) # same data, but stored as miVariable x <- miVariable(c(1,2,NA,NA), list(Imp1 = c(3,4), Imp2 = c(4,2))) x # Use of the subset argument (otherwise like x just above): miVariable(c(1,2,NA,NA), subset = T, list(Imp1 = c(99,99,3,4), Imp2 = c(99,99,4,2))) # as.miVariable does not always return an miVariable object y <- as.miList(x) as.miVariable(y) # an miVariable y[[1]] <- 1:6 # Now first imputation is longer as.miVariable(y) # miList (couldn't convert to miVariable)