Publication
A non-parametric method of distribution estimation for univariate data is presented. The idea is to adapt the smoothing spline procedure used in regression to the estimation of distributions via a scatterplot smoothing of theempirical distribution function. An explicit formula for the estimator is obtained by minimizing a penalized weighted sum of squares. The issue of monotonicity of the resulting function is discussed in detail and the estimator's large sample properties are studied.
Daniel Kuhn, Yves Rychener, Viet Anh Nguyen