If there was only one other matching case, its response to the question of interest was
donated to the case with missing data. If there was more than one matching case, one was chosen at random as the donor, provided one condition was met. The condition was that, for potential donor cases, the variance of scores on the question of interest must be below the variance for the sample at large. We required that the variance for
potential donors be less than 70% of that for the full sample. Imputation permitted us to maximize retention of students thereby minimizing the introduction of bias while having confidence that scale scores adequately reflected student knowledge, attitudes and experiences.