Chapter 45: The MIANALYZE Procedure

Overview

The MIANALYZE procedure combines the results of the analyses of imputations and generates valid statistical inferences. Multiple imputation provides a useful strategy for analyzing data sets with missing values. Instead of filling in a single value for each missing value, Rubin s (1976; 1987) multiple imputation strategy replaces each missing value with a set of plausible values that represent the uncertainty about the right value to impute. You can implement the strategy with two SAS procedures: PROC MI, which generates imputed data sets, and PROC MIANALYZE, which combines the results of analyses carried out on the data sets.

The analyses of imputations are obtained by using standard SAS procedures (such as PROC REG) for complete data. No matter which complete-data analysis is used, the process of combining results from different imputed data sets is essentially the same and results in valid statistical inferences that properly reflect the uncertainty due to missing values.

The MIANALYZE procedure reads parameter estimates and associated standard errors or covariance matrix that are computed by the standard statistical procedure for each imputed data set. The MIANALYZE procedure then derives valid univariate inference for these parameters. With an additional assumption about the population between and within imputation covariance matrices, multivariate inference based on Wald tests can also be derived.

For some parameters of interest, you can use TEST statements to test linear hypotheses about the parameters. For others, it is not straightforward to compute estimates and associated covariance matrices with standard statistical SAS procedures. Examples include correlation coefficients between two variables and ratios of variable means. These special cases are described in the Examples of the Complete-Data Inferences section on page 2628.

In SAS 9, the VAR statement has been replaced by the MODELEFFECTS statement to accommodate various effects to be analyzed . A STDERR statement has been added to specify the standard errors when both parameter estimates and associated standard errors are stored as variables in the same data set.

Also, an experimental CLASS statement has been added to specify classification variables in the MODELEFFECTS statement. The CLASS statement must be used in conjunction with the MODELEFFECTS statement.