Nonparametric analysis of randomized experiments with missing covariate and outcome data

Analysis of randomized experiments with missing covariate and outcome data is problematic, because the population parameters of interest are not ident ified unless one makes untestable assumptions about the distribution of the missing data. This article shows how population parameters can be bounded without making untestable distributional assumptions. Bounds are also deriv ed under the assumption that covariate data are missing completely at rando m. In each case the bounds are sharp; they exhaust all of the information a vailable given the data and the maintained assumptions. The bounds are illu strated with applications to data obtained from a clinical trial and data r elating family structure to the probability that a youth graduates from hig h school.