1
$\begingroup$

Let $A_1,A_2,\dots,A_n$ be events on a probability space. For $0 \leq k \leq n$ let \begin{equation*} S_k=\sum_{1 \le {i_1}<{i_2}<\cdots<{i_k} \leq n} P(A_{i_1} \cap \cdots \cap A_{i_k}). \end{equation*} This is the $k$-th binomial moment of the number $m_n$ of those $A$s which occur.

Question: What bounds are known for $S_k$ in terms of $S_0,S_1,\dots,S_{k-1}$?

Note: Bonferroni-type inequalities give bounds for $P(m_n \geq t)$ in terms of linear combinations of the $S_i$ (for example Galambos, "Bonferroni Inequalities", Annals of Probability, Vol. 5 (1977), 577--581). It is possible to use these to give bounds for $S_k$ in terms of $S_0,S_1,\dots,S_{k-1}$ and $P(m_m \geq t)$ and use the fact that $0 \leq P(m_n \geq t) \leq 1$ to deduce bounds on $S_k$ in terms of $S_0,S_1,\dots,S_{k-1}$. It seems to me however that a lot is wasted that way. I am wondering if there are known ways of obtaining sharper bounds.

$\endgroup$
2
$\begingroup$

There are better methods using tree and graphs, for example, "Probability Bounds with Cherry Trees" in Math. of Oper. Research in 2001.

$\endgroup$
1
  • $\begingroup$ Thanks! This paper looks interesting; I'll check it out. $\endgroup$ – Siksek Mar 3 '13 at 22:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.