> Pat, > Why not just do the bootstrap from the missing data set directly and obtain > direct ML estimates (under MAR assumptions) from each of your bootstrap > samples?
Because of the measurement aspects. I'm using an indicator for measurement (i.e., Y=1 if any of p, q, r = 1, and 0 otherwise). Imputation has seemed to be the best way to get an unbiased assessment of Y when p, q, and r are missing in arbitrary patterns -- I impute the component variables and derive the indicator separately by imputation. If anyone's got an easier solution for handling the measurement problem, I'm all ears. Thanks, Pat -- Patrick S. Malone, Ph.D., Research Scholar Duke University Center for Child and Family Policy Durham, North Carolina, USA e-mail: [email protected] http://www.duke.edu/~malone/
