Missing data
Missing data were handled using the multiple imputation
approach, which results in less bias than alternative procedures
such as list-wise deletion or mean substitution
[24,25]. Multiple imputation was implemented using a SAS
procedure PROC MI with the Markov Chain Monte Carlo
(MCMC) method [26]. Parameter estimates were combined
and statistical inferences were made using another SAS
procedure, PROC MIANALYZE [26]. Twenty imputed
data sets were generated to ensure a relative efficiency [27].
All results reported below are the average [24,25] of the
results based on 20 imputed data sets.