The infimum of a moment generating function

550 Views Asked by At

Studying probability theory using Grimmett and Stirzaker's book and their solution for this question was (predictably) sparse, so I'm hoping someone else can explain it:

For $X$ a random variable with moment generating function $M_{X}$, show that:

$P(X\geq x) \leq \inf_{t \geq 0} \left \{e^{-tx}M_{X}(t) \ \right \} $

1

There are 1 best solutions below

1
On BEST ANSWER

They used Chernoff's bound. Which itself is derived from Markov's bound.
Markov's bound is (for non-negative random variable $X$) $$\mathbb{P}(X>x)\le \frac {\mathbb{E}(X)}{x}$$ Then
$$\mathbb{P}(e^{tX}>e^{tx})\le \frac {\mathbb{E}(e^{tX})}{e^{tx}}= M_X(t)e^{-tx} $$ Since this is true for all $t > 0 $ (since when $t<0$ the inequality inside the probability gets reversed), you can choose $t$ such that the RHS is minimized.