Suppose we have two events $A, B \subset \Omega$ in a probability space $(\Omega, \mathcal{F}, \mathbf{P})$ and we want to know the probability of $A \cap B$.

If $A$ and $B$ are independent, of course $\mathbf{P}(A \cap B) = \mathbf{P}(A)\mathbf{P}(B)$.

However, what can we say without independence? It seems that with Cauchy-Schwarz we have that $$ \mathbf{P}(A \cap B) = \mathbf{E} 1_A 1_B \leq \sqrt{\mathbf{P}(A) \mathbf{P}(B)}. $$ Is this a good bound? Is it useful? That's a soft question, but I guess I am wondering if it is really useful at all.


This bound is not really useful because $\min({\bf P}(A), {\bf P}(B))$ is better. Note that $$\sqrt{{\bf P}(A) {\bf P}(B)} \ge \min({\bf P}(A),{\bf P}(B)) \ge {\bf P}(A \cap B)$$

  • $\begingroup$ Fair. Recording another thought here, in case others find it useful: you might ask, okay what if $1 < p < \infty$, $q$ solves $1/p + 1/q = 1$? Holder's inequality also gives you the bound $\mathbf{P}(A \cap B) \leq \mathbf{P}(A)^{1/p} \mathbf{P}(B)^{1/q}$, but a similar argument above holds; this is also not useful. $\endgroup$ – Drew Brady Jan 3 at 6:47

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.