Abstract
Power of alternative tests for two paired samples with missing data. When a research design that will use a test of means difference for matched samples is to be applied sometimes part of the observations on either of the variables are missing. The usual way to face those situations is the supression of the incomplete subjects or observations. In the present paper the consequences of the supression are analyzed, other alternatives are reviewed and a simulation study is reported. In the simulation the consequences of those alternatives over the probabilities of Type I and Type II errors are studied. Some practical recommendations for these situations are also included.
