Introduction
Conditional Probability is the concept that addresses a fundamental question: how should we update our beliefs in light of the evidence we observe
In real world, a useful perspective is that all probabilities are conditional since whether or not it's written explicitly, there is always background knowledge built into every probability
Definition
If and are events with , then the conditional probability of given , denoted by , is defined as
Here is the event whose uncertainty we want to update, and is the evidence we observe (or want to treat as given). We call the prior probability of and the posterior probability of ("prior" means before updating based on the evidence, and "posterior" means after updating based on the evidence)
It's natural to conclude that for any event , since upon observing that has occurred, out updated probability for is
Bayes' rule and the law of total probability
Probability of the intersection of two events
At first sight this theorem may not seem very useful: it is the definition of conditional probability, just written slightly differently. But the theorem is actually very useful, since it often turns out to be possible to find conditional probabilities without going back to the definition
Probability of the intersection of events
For any events with
In fact, this is theorems in one, since we can permute however we want without affecting the left-hand side.
Bayes' rule
See also Bayes' Theorem
Law of total probability
Let be a partition of the sample space (i.e., the are disjoint events and their union is ), with for all . Then
Conditional probabilities are probabilities
For conditional probabilities, the laws of probability operate as well, in particular when we condition on an event :
- Conditional probabilities are in
- If are disjoint, then
- Inclusion-exclusion:
Baye's rule with extra conditioning
Consider as a probability function which assigns probabilities in accordance with the knowledge that has occurred.
LOTP with extra conditioning
Independence of events
Independence of two events
Events and are independent if
If , this is equivalent to
and also equivalent to
Note that independence is a symmetric relation: if is independent of , then is independent of
Warning
Independent is completely different from disjointness
If and are independent, then and are all independent
Independence of three events
Events are said to be independent if all of the equations below hold
If the first three conditions hold, we say that , , and are pairwise independent. Pairwise independence does not imply independence, and neither vice-versa.
Independence of events
We say events are independent if all the subsets of satisfying , where
Conditional independence
Events and are said to be conditional independence given if
Tip
- Two events can be conditionally independent given , but not independent given .
- Two events can be conditionally independent given , but not independent.
- Two events can be independent, but not conditionally independent given .
Coherency of Bayes' rule
An important property of Bayes' rule is that it is coherent: if we receive multiple pieces of information and wish to update our probabilities to incorporate all the information, it does not matter whether we update sequentially, taking each piece of evidence into account one at time, or simultaneously, using all the evidence at once.