A note on estimating a non-increasing density in the presence of selection bias