A note on estimating a smooth monotone regression by combining kernel and density estimates