In probability theory, the multidimensional Chebyshev's inequality[1] is a generalization of Chebyshev's inequality, which puts a bound on the probability of the event that a random variable differs from its expected value by more than a specified amount.
Let
X
N
\mu=\operatorname{E}[X]
V=\operatorname{E}[(X-\mu)(X-\mu)T].
If
V
t>0
\Pr\left(\sqrt{(X-\mu)TV-1(X-\mu)}>t\right)\le
N | |
t2 |
Since
V
V-1
y=(X-\mu)TV-1(X-\mu).
Since
y
\Pr\left(\sqrt{(X-\mu)TV-1(X-\mu)}>t\right)=\Pr(\sqrt{y}>t)=\Pr(y>t2) \le
\operatorname{E | |
[y]}{t |
2}.
Finally,
\begin{align} \operatorname{E}[y]&=\operatorname{E}[(X-\mu)TV-1(X-\mu)]\\[6pt] &=\operatorname{E}[\operatorname{trace}(V-1(X-\mu)(X-\mu)T)]\\[6pt] &=\operatorname{trace}(V-1V)=N \end{align}.
lX
lX
Suppose that is of "strong order two", meaning that
\operatorname{E}\left(\|
2 | |
X\| | |
\alpha |
\right)<infty
for every seminorm . This is a generalization of the requirement that have finite variance, and is necessary for this strong form of Chebyshev's inequality in infinite dimensions. The terminology "strong order two" is due to Vakhania.[4]
Let
\mu\inlX
\sigmaa:=\sqrt{\operatorname{E}\|X-
2} | |
\mu\| | |
\alpha |
be the standard deviation with respect to the seminorm . In this setting we can state the following:
General version of Chebyshev's inequality.
\forallk>0: \Pr\left(\|X-\mu\|\alpha\gek\sigma\alpha\right)\le
1 | |
k2 |
.
Proof. The proof is straightforward, and essentially the same as the finitary version. If, then is constant (and equal to) almost surely, so the inequality is trivial.
If
\|X-\mu\|\alpha\gek
2 | |
\sigma | |
\alpha |
then, so we may safely divide by . The crucial trick in Chebyshev's inequality is to recognize that
1=\tfrac{\|X-
2}{\|X | |
\mu\| | |
\alpha |
-
2} | |
\mu\| | |
\alpha |
The following calculations complete the proof:
\begin{align} \Pr\left(\|X-\mu\|\alpha\gek\sigma\alpha\right)&=\int\Omega
1 | |
\|X-\mu\|\alpha\gek\sigma\alpha |
d\Pr\\ &=\int\Omega\left(
| |||||||||
|
\right) ⋅
1 | |
\|X-\mu\|\alpha\gek\sigma\alpha |
d\Pr\\[6pt] &\le\int\Omega\left(
| |||||||||
|
\right) ⋅
1 | |
\|X-\mu\|\alpha\gek\sigma\alpha |
d\Pr\\[6pt] &\le
1 | ||||||||||||
|
\int\Omega\|X-
2 | |
\mu\| | |
\alpha |
d\Pr&&
1 | |
\|X-\mu\|\alpha\gek\sigma\alpha |
\le1\\[6pt] &=
1 | ||||||||||||
|
\left(\operatorname{E}\|X-
2 | |
\mu\| | |
\alpha |
\right)\\[6pt] &=
1 | ||||||||||||
|
\left
2 | |
(\sigma | |
\alpha |
\right)\\[6pt] &=
1 | |
k2 |
\end{align}