Linear boundary kernels for bivariate density estimation

We propose a new linear boundary kernel for bivariate density estimation on a compact region. This kernel is simple to implement even for irregular boundaries, and reduces the boundary bias to the same asymptotic order as for an interior point.