Let $(\Omega, M, P)$ be a probability space and let $O_1, O_2$ be independent events. Does the sigma algebras generated by $F_1 := \sigma(\{O_1\})$ and $F_2 := \sigma(\{O_1,O_2\})$ are independent? Here $O_1,O_2$ are not disjoint and $1 > P(O_1),P(O_2) > 0$.

The sigma algebra generated by $O_1$ is $\{\emptyset, O_1, O_1^\complement, \Omega\}$ where $O_1^\complement$ is taken as the complement relative to $\Omega$. We have,

$$F_1 = \{\emptyset, O_1, O_1^\complement, \Omega\}.$$

On the other hand, the sigma algebra generated by $F_2$ is given by $$F_2=\{\emptyset, \Omega, O_1,O_2,O_1^\complement,O_2^\complement,O_1\cup O_2, O_1 \cap O_2,O_1^\complement \cap O_2^\complement, O_1^\complement\cup O_2^\complement\}.$$

Now, on the other hand, I know that $O_1^\complement, O_2$, are independent and that $O_1^\complement, O_2^\complement$ are also independent. I think they are not independent, because for instance, $$P(O_1\cap(O_1^\complement \cap O_2^\complement)) = P((O_1\cap O_1^\complement) \cap O_2^\complement) = P(\emptyset) = 0 \neq P(O_1)P(O_1^\complement) P(O_2^\complement).$$

Does this shows that the sigma algebras generated by $F_1, F_2$ are not independent?


Any two sigma algebras that have an event $E$ in common with $0<P(E)<1$ are not independent. Why? Let $E\in\mathcal F_1$ and $E\in\mathcal F_2.$ Then for $\mathcal F_1$ and $\mathcal F_2$ to be independent we would need $$ P(E\cap E) = P(E)P(E)$$ but $$ P(E\cap E) = P(E)$$ so we have $$ P(E) = P(E)^2$$ which means $P(E) = 0$ or $1.$

  • $\begingroup$ Thank you very much, I was looking for a simple proof like this one. $\endgroup$ – okie Feb 20 '17 at 5:11
  • $\begingroup$ @richitesenpai no problem. Check my edit though, there is one corner case where they can still be independent in a strict reading of your question. $\endgroup$ – spaceisdarkgreen Feb 20 '17 at 5:16
  • $\begingroup$ Yes, actually, I know the probability of both sets, which are less than 1. I wanted to see a general case. Will edit the post and thanks again. $\endgroup$ – okie Feb 20 '17 at 5:19

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.