Use of multiple imputation to replace missing outcomes in clinical research is a relatively new approach. Thus, there is little information published on how to obtain and combine the types of estimates needed to use multiple imputation in analysis of results from many clinical trials. This presentation summarizes procedures developed, or adapted from other sources, using SAS®, through the stages of preparation of the data for imputation, processing after imputation, setting up output of SAS PROCs (e.g., GLM, REG, MIXED) for processing by PROC MIANALYZE, and other routines we developed for output not processed by PROC MIANALYZE. This paper does not report on the efficiency or unbiasedness of the results. Rather, we focus on providing guidance on the use of SAS® by clinical investigators in order to yield meaningful results from analysis of a multiply-imputed dataset.
conference:
Paper Type:
Paper
