The MI Procedure

(View the complete code for this example.)

This example uses FCS regression methods to impute values for all continuous variables in a data set that has an arbitrary missing pattern.

The following statements invoke the MI procedure and impute missing values for the Fitness1 data set:

proc mi data=Fitness1 seed=1213 nimpute=pctmissing(min=5 max=20)
        mu0=50 10 180 out=outex6;
   fcs nbiter=20 reg(Oxygen/details);
   var Oxygen RunTime RunPulse;
run;

The NIMPUTE=PCTMISSING option uses the percentage of the incomplete cases as the number of imputations. The MIN=5 (which is the default) and MAX=20 options restrict the number of imputations to be in the range of 5 to 20. That is, 5 imputations are generated if the percentage of the incomplete cases is less than 5, and 20 imputations are generated if this percentage is greater than 20.

The FCS statement requests multivariate imputations by FCS methods, and the NBITER=20 option (which is the default) specifies the number of burn-in iterations before each imputation.

The "Model Information"  table in Output 76.6.1 describes the method and options used in the multiple imputation process.

The "FCS Model Specification"  table in Output 76.6.2 describes methods and imputed variables in the imputation model. With the REG(OXYGEN) option in the FCS statement, the procedure uses the regression method to impute variable Oxygen. By default, the regression method is also used to impute variables RunTime and RunPulse.

The "Missing Data Patterns" table in Output 76.6.3 lists distinct missing data patterns with corresponding frequencies and percentages.

For the NIMPUTE=PCTMISSING option, the percentage of the incomplete cases, 10/31 = 32.3%, is used as the number of imputations. But the number 33 (after rounding up) is greater than 20 (as specified in the MAX= option), so only 20 imputations are generated.

When you specify the DETAILS option in REG(OXYGEN/DETAILS), the parameters that are used in each imputation for Oxygen are displayed in Output 76.6.4.

The following statements list the first 10 observations of the data set Outex6 in Output 76.6.5. Note that all missing values of all variables are imputed.

proc print data=outex6(obs=10);
   title 'First 10 Observations of the Imputed Data Set';
run;

After the completion of the specified four imputations, the "Variance Information" table in Output 76.6.6 displays the between-imputation variance, within-imputation variance, and total variance for combining complete-data inferences. 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 76.6.7 displays a 95% mean confidence interval and a t statistic with its associated p-value for each of the hypotheses requested with the MU0= option.