The
cholesterol
and
cholesterolImpExample
data frames have 28 rows and 3 columns.
Cholesterol measurements were recorded for 28 patients
treated for heart attacks at a Pennsylvania medical center.
These data frames contain the following columns:
chol14
has 9 missing values. In the
cholesterolImpExample
data frame this variable is an
miVariable
object, containing
the original data and 5 sets of multiple imputations for the missing values.
Ryan, B. F. and Joiner, B. L. (1994) Minitab Handbook (Third Edition). Wadsworth, Belmont, CA.
Schafer, J. L. (1997),
Analysis of Incomplete Multivariate Data ,
Chapman & Hall, London.
Schimert, J., Schafer, J. L., Clarkson, D. B., Fraley, C., Hesterberg, T., (2000)
Analyzing Data with Missing Values in S-PLUS ,
Insightful Corporation, Seattle, Washington
(This manual is available on-line as file Missing.pdf),
Chapter 9.
# Estimate parameters using the EM algorithm: mdGauss(object = cholesterol, na.proc = "em") # Create multiple imputations. # Create overdispersed starting values using bootstrap samples of size n/2 start <- list() # Store starting values in a list set.seed(57) # Set the random number seed for reproducibility for(i in 1:5) start[[i]] <- paramIter(emGauss(cholesterol, subset=sample(1:28, size=14, T), prior="ml", control= list(trace= F))) # Generate imputations using those starting values cholesterolImpExample <- impGauss(cholesterol, prior="non", start=start, control=list(niter=50, seed=21))