| Previous: How Do I Use | Up: Frequently Asked Questions | Next: Why Preprocessing? |
MATCHIT requires complete data sets, with no missing values (other than potential outcomes of course). If there are missing values in the data set, imputation techniques should be used first to fill in (``impute'') the missing values (both covariates and outcomes), or the analysis should be done using only complete cases (which we do not in general recommend). For imputation software, see Amelia at (http://gking.harvard.edu/stats.shtml) or other programs at http://www.multiple-imputation.com. For more information on missing data and imputation methods, see King et al. (2001).