The MI Procedure

This example uses FCS methods to impute missing values in both continuous and CLASS variables in a data set with an arbitrary missing pattern. The following statements invoke the MI procedure and impute missing values for the Fish3 data set:

proc mi data=Fish3 seed=1305417 out=outex7;
   class Species;
   fcs nbiter=10 discrim(Species/details) reg(Height/details);
   var Species Length Height Width;
run;

The DISCRIM option uses the discriminant function method to impute the classification variable Species, and the REG option uses the regression method to impute the continuous variable Height. By default, the regression method is also used to impute other continuous variables, Length and Width.

The “Model Information”  table in Output 57.7.1 describes the method and options used in the multiple imputation process.

The “FCS Model Specification”  table in Output 57.7.2 describes methods and imputed variables in the imputation model. The procedure uses the discriminant function method to impute the variable Species, and the regression method to impute other variables.

The “Missing Data Patterns” table in Output 57.7.3 lists distinct missing data patterns with corresponding frequencies and percentages.

With the specified DETAILS option for variables Species and Height, parameters used in each imputation for these two variables are displayed in the “Group Means for FCS Discriminant Method” table in Output 57.7.4 and in the “Regression Models for FCS Method” table in Output 57.7.5.

The following statements list the first 10 observations of the data set outex7 in Output 57.7.6:

proc print data=outex7(obs=10);
   title 'First 10 Observations of the Imputed Data Set';
run;

After the completion of five imputations by default, the “Variance Information” table in Output 57.7.7 displays the between-imputation variance, within-imputation variance, and total variance for combining complete-data inferences for continuous variables. The relative increase in variance due to missingness, the fraction of missing information, and the relative efficiency for each variable are also displayed. These statistics are described in the section Combining Inferences from Multiply Imputed Data Sets.

The “Parameter Estimates” table in Output 57.7.8 displays a 95% mean confidence interval and a t statistic with its associated p-value for each of the hypotheses requested with the default MU0=0 option.