Almost all economic data sets are discretized or rounded to some extent. This paper proposes a regression and a density estimator that work especially well when the data is very discrete. The estimators are a weighted average of the data, and the weights are composed of cubic B-splines. Unlike...