Multiple Imputation of Missing Data Using SAS by Patricia Berglund & Steven G. Heeringa

Author:Patricia Berglund & Steven G. Heeringa [Berglund, Patricia]
Language: eng
Format: azw3
Publisher: SAS Institute
Published: 2014-07-01T04:00:00+00:00

Output 5.3: Variance Information and Parameter Estimates from PROC MI

Outputs 5.2 and 5.3 display the Model Information, Variance Information, and Parameter Estimates tables produced by PROC MI.

The Model Information table details the method used (MCMC with a single chain), use of the EM algorithm for initial estimates for the chain, the default Jeffreys prior for the multivariate normal parameters, 200 burn-in iterations, 100 iterations separating each repetition imputation in the single chain, a seed value of 192, and number of imputation repetitions (10).

The Variance Information table provides information on the estimates of the MI variance components for the estimated mean for player salaries and number of strike outs. The initial columns provide the estimated between imputation, within imputation, and total variance for an MI estimate of the mean of player salaries. Additional MI statistics included in this output table are the Relative Efficiency (RE= 0.99 for both imputed variables), the Relative Increase in Variance (RIV=0.03 and 0.01) and the Fraction of Missing Information (FMI=0.03 and 0.01).

The Parameter Estimates table in the PROC MI output provides the MI estimates of the mean for SALARY (1267.42) and STRIKE_OUTS (56.67), the standard errors and 95% confidence limits along with other descriptive statistics from the multiply imputed data set. The variability or uncertainty introduced by the imputation process is factored into these estimates through use of the within and between imputation variances. The default Mu0=0 and Student t statistics for testing H0: Mean=Mu0, along with the p values, specify and test the null hypothesis that the means for salary and strike outs are equal to 0. In both cases, they are significantly different from 0 at p values of <.0001.

Figure 5.1: Trace Plot for Salary

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.