All regular [[Probability Axioms]] apply to [[Conditional Probability]] as well.
- *Non-negativity:* All conditional probabilities are non-negative. For that it just needs to be ensured that the condition has a positive probability $\mathbf P(B) >0$.
$ \mathbf P(A \vert B) \ge 0 $
- *Normalization:* Within the conditioned [[Sample Space]], all probabilities sum to $1$.
$ \mathbf P(\Omega \vert B) = \frac{\mathbf P(\Omega \cap B)}{\mathbf P(B)} = \frac{\mathbf P(B)}{\mathbf P(B)}=1 $
- *Additivity:* When two [[Events]] are disjoint $A \cap C= \emptyset$, they are also disjoint in any possible conditional sample space.
$ \mathbf P(A \cup C \vert B)=\mathbf P(A \vert B) + \mathbf P(C\vert B) $
Let us prove that additivity also works in the conditional case:
$
\begin{align}
\mathbf P(A \cup C \vert B)
&= \frac{\mathbf P\big((A \cup C) \cap B\big)}{\mathbf P(B)} \tag{1}\\[6pt]
&= \frac{\mathbf P(A \cap B) \cup \mathbf P (B \cap C)}{\mathbf P(B)} \tag{2}\\[6pt]
&= \frac{\mathbf P(A \cap B)}{\mathbf P(B)} \cup \frac{\mathbf P(B \cap C)}{\mathbf P(B)} \tag{3}\\[6pt]
&=\mathbf P(A \vert B) + \mathbf P(C \vert B) \tag{4}
\end{align}
$
where:
- (1) We express the conditional probability in the form of Bayes theorem.
- (3) The numerator can be split into two terms.