Togaware DATA MINING
Desktop Survival Guide
by Graham Williams
Google

Imputation

Multiple imputation (MI) is a general purpose method for handling of missing data. The basic idea is: Impute missing values using an appropriate model that incorporates random variation; Do this $m$ times (often 3-5 times) to obtain $m$ datasets, all with no missing values; Do the intended analysis on each of these datasets; Gert the average values of the parameter estimates across the $m$ samples to have a single point estimate; Calculate standard errors by firstly averaging the squared standard errors of the $m$ estimates and calculating the variance of the $m$ parameter estimates across samples, and then combine these in some way.

There are a number of R packages for imputation.



Copyright © 2004-2006 [email protected]
Support further development through the purchase of the PDF version of the book.
Brought to you by Togaware.