The existence of the least favorable noise

Abstract: Suppose that a random variable

X

of interest is observed. This paper concerns "the least favorable noise"

h a t Y_{e p s i l o n}

, which maximizes the prediction error

E [X - E [X | X + Y]]^{2}

(or minimizes the variance of

E [X | X + Y]

) in the class of

Y

with

Y

independent of

X

and

m a t h r m v a r Y l e q e p s i l o n^{2}

. This problem was first studied by Ernst, Kagan, and Rogers ([3]). In the present manuscript, we show that the least favorable noise

h a t Y_{e p s i l o n}

must exist and that its variance must be

e p s i l o n^{2}

. The proof of existence relies on a convergence result we develop for variances of conditional expectations. Further, we show that the function

i n f_{m a t h r m v a r Y l e q e p s i l o n^{2}}, m a t h r m v a r, E [X | X + Y]

is both strictly decreasing and right continuous in

e p s i l o n

.

Recommendations

Cites Work

Cited In (2)

This page was built for publication: The existence of the least favorable noise