miMeanSEDefault(estimate, se, df = NA, n = <see below>, sse = F, cov)
estimate, but containing complete-data standard errors
for the estimates.
df=Inf
(infinity) to obtain degrees of freedom for problems
in which complete-data inferences would use normal or chi-square
rather than t or F distributions.
df plus
the number of parameters in
estimate.
TRUE then
se is assumed to contain squared standard
errors (variances) rather than standard errors.
estimate, but containing complete-data covariance
matrices for the estimates, whose diagonal elements are
squared standard errors. If
cov is supplied then
se is
ignored.
This function performs core calculations for combining estimates
and standard errors, and is called my
miMeanSE and its
auxiliary function, including
miMeanSEAux and its methods, and
miMeanSEList
and
miMeanSEMatrix.
If
df or
n contain non-identical imputations, warnings
are issued and certain averages used. If either of these
are not supplied, or contain missing values (
NA), then
some components of the output will contain missing values.
fit <- miEval(lm(chol14~., data = cholesterolImpExample))
sumfit <- miEval(summary(fit))
coefs <- miEval(coef(sumfit))
coefs[[1]] # note estimates and std errors
miMeanSEDefault(miEval(coefs[,1]), miEval(coefs[,2]))
# May use miMeanSE as a front end
miMeanSE(miEval(coefs[,1]), miEval(coefs[,2]))
# May pass additional information, e.g. residual degrees of freedom
miMeanSE(miEval(coefs[,1]), miEval(coefs[,2]),
df = nrow(cholesterolImpExample) - 3)