> Hello, > I have multiply imputed data sets. Each data set will be analyzed using > multilevel models (SAS Proc MIXED). Is it legitimate (statistically > defensible) to combine the data sets into one "long" data set, add > imputation number as a new (highest) level in the multilevel model, > analyze the data with a single model run? Would I need to down weight the > observations in each multiply imputed data set? Any ideas or references > would be appreciated. Thanks. > Ted > > Mr. Theodore J. Thompson [EMAIL PROTECTED] > Chief, Statistics Section > Division of Diabetes Translation > Centers for Disease Control and Prevention > 4770 Buford Highway > Mailstop K-68 > Atlanta, Georgia 30341 > 770-488-1270 > > "Randomization by cluster accompied by an analysis appro- > priate to randomization by individual is an exercise in self- > deception, however, and should be discouraged." > Cornfield > (1978) >
