Linear independence check can be done as: given vectors $v_1, v_2, \dots v_n$, if there exists $a_1, a_2, \dots a_n$, not all zero, such that $$a_1 v_1 + a_2 v_2 + \dots + a_n v_n = 0$$, then $v_i$ are not linear independent; otherwise, $v_i$ are linear independent.

This check is quite straight-forward as it will lead to solve a $n\times n$ linear equation, or computing the det of a $n\times n$ matrix.

Probability independence check is a bit different. It's defined as:

Given a probability triple $(\Omega, F, P)$, a collection of events $\{A_1, A_2, \dots, A_n \}$ is independent if for all $k \in N$, and all possible $\{i_1, i_2, \dots, i_k\}$, $$P(A_{i_1}\cap A_{i_2} \cap \dots \cap A_{i_k})=P(A_{i_1})P(A_{i_2}) \dots P(A_{i_k})$$

This is quite troublesome as there are so many combinations.

Is there a similar way to check the independent of a collection of events? Or even better is it possible to convert the problem to check of vectors' linear independence?

  • $\begingroup$ I'm afraid you are stuck with troublesome. The fact that each situation uses the word "independent" doesn't really help with the arguments. $\endgroup$ – Ethan Bolker Jul 4 at 18:31
  • $\begingroup$ @EthanBolker i'm not wishing independence always brings the same check; I'm just wishing there's an easy way to check Probability Independence. is that possible? $\endgroup$ – athos Jul 4 at 18:32
  • $\begingroup$ I don't think so... $\endgroup$ – Ethan Bolker Jul 4 at 18:38

As noted by Ethan Bolker, there is no relation between the two meanings of "independent", and no shortcut in general.

I might note that it is entirely possible that of these $2^n$ equations, $2^n - 1$ are true, but one is false.

However, in practice one almost never actually checks all these equations. Rather, some assumptions are made in setting up a probability model, as a consequence of which we can deduce the independence. Thus if the $A_i$ depend only on different parts of an experiment that have no physical connection to each other, then they are independent.


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.