Nonparametric estimation of possibly similar densities

A class of nonparametric methods is developed to estimate a set of possibly similar densities that offers greater efficiency if they are similar while seemingly not losing any if they are not. Theoretical properties and finite sample performance are promising.