Extension of Chebyshev Inequality
Theorem1
Let be a random variable with expected value and variable .
Let be a probability distribution function.
Let be .
Then, for any r and k>0,
, inequality in continuous case coincides with Chebyshev inequality.
Theorem 2
Let be a random variable, be a probability mass function.
Let m be .
Let
Then, for any r and k>0,
Proof of Theorem 1
Consider the function
.
Here,
By definiton of m(k), eq(1)
By deforming
eq(2)
By definition of Pr, , and
is convex function for , and .
We can apply Jesen's inequality for eq(2).
.
Here,
Then we have
.
By the result of eq(1),
Deforming this inequality, we have
.
Proof of Theorem 2
We put if
U(X) is continuous uniform distribution.
Then,
.
is the variance of , and is the variance of .
We put , .
We have
.