The MIANALYZE Procedure

Example 62.7 Reading Logistic Model Results from a PARMS= Data Set

This example creates data sets that contains parameter estimates computed by a logistic regression analysis for a set of imputed data sets. These estimates are then combined to generate valid statistical inferences about the model parameters.

The following statements use PROC LOGISTIC to generate the parameter estimates for each imputed data set:

proc logistic data=outfish2;
   class Species;
   model Species= Length Width / covb;
   by _Imputation_;
   ods output ParameterEstimates=lgsparms;
run;

The following statements display (in Output 62.7.1) the output logistic regression coefficients from PROC LOGISTIC for the first two imputed data sets:

proc print data=lgsparms (obs=8);
   title 'LOGISTIC Model Coefficients (First Two Imputations)';
run;

Output 62.7.1: PROC LOGISTIC Model Coefficients

LOGISTIC Model Coefficients (First Two Imputations)

Obs _Imputation_ Variable DF Estimate StdErr WaldChiSq ProbChiSq _ESTTYPE_
1 1 Intercept 1 0.1637 1.8405 0.0079 0.9291 MLE
2 1 Length 1 1.4543 0.5167 7.9231 0.0049 MLE
3 1 Width 1 -10.2950 3.4860 8.7216 0.0031 MLE
4 2 Intercept 1 0.6473 1.9003 0.1160 0.7334 MLE
5 2 Length 1 1.2831 0.4778 7.2123 0.0072 MLE
6 2 Width 1 -9.2991 3.2187 8.3469 0.0039 MLE
7 3 Intercept 1 -0.0408 1.8535 0.0005 0.9824 MLE
8 3 Length 1 0.9208 0.3978 5.3564 0.0206 MLE


The following statements displays the covariance matrices associated with parameter estimates derived from the first two imputations in Output 62.7.2:

The following statements use the MIANALYZE procedure with input PARMS= data set:

proc mianalyze parms=lgsparms;
   modeleffects Intercept Length Width;
run;

The Variance Information table in Output 62.7.2 displays the between-imputation, within-imputation, and total variances for combining complete-data inferences.

Output 62.7.2: Variance Information

The MIANALYZE Procedure

Variance Information
Parameter Variance DF Relative
Increase
in Variance
Fraction
Missing
Information
Relative
Efficiency
Between Within Total
Intercept 0.125100 3.174905 3.325025 1962.3 0.047283 0.046120 0.990860
Length 0.039992 0.201496 0.249486 108.11 0.238169 0.206894 0.960265
Width 1.895087 9.030840 11.304945 98.85 0.251815 0.216847 0.958433


The Parameter Estimates table in Output 62.7.3 displays the combined parameter estimates with associated standard errors.

Output 62.7.3: Parameter Estimates

Parameter Estimates
Parameter Estimate Std Error 95% Confidence Limits DF Minimum Maximum Theta0 t for H0:
Parameter=Theta0
Pr > |t|
Intercept 0.073984 1.823465 -3.5021 3.65012 1962.3 -0.208872 0.647303 0 0.04 0.9676
Length 1.191908 0.499485 0.2019 2.18196 108.11 0.920752 1.454324 0 2.39 0.0188
Width -8.499960 3.362283 -15.1716 -1.82834 98.85 -10.294965 -6.703819 0 -2.53 0.0131