PROC IRT handles missing values differently for different tasks. During model estimation and subject scoring, observations with missing values are still used, in the sense that the nonmissing values can still contribute to the estimation or scoring. When calculating the polychoric correlation for a pair of variables, PROC IRT does pairwise deletion, in which observations that have valid data about the corresponding pair of variables are used. When calculating the item statistics table, PROC IRT uses listwise deletion, in which observations with missing values for any variables in the analysis are omitted from the computations.