Probability calculation from distributions.

32 Views Asked by At

The scenario below will help me understand how to apply the law of total probability to nonbinary examples. It is inspired by an exercise from a master's program in AI using Bayesian Networks I'm currently under. The scenario is fictional and it was made purely to show how to use the theorem with a more complicated example. The law of total probability is a requirement for understanding the probability of dependent variables and events.

Let $A$ and $B$ be independent identically distributed random variables. $A$ and $B$ can assume values from $[0, 1, 2]$ with probabilities $[0.25, 0.25, 0.50]$ respectively. Now, let event $E$ be dependent on $A$ and $B$, so the probability of $E$ being $X$, $Y$ or $Z$ can be summarized as follows:

\begin{array}{c|lll} E(B-A) & X & Y & Z \\\hline 0 & 0.15 & 0.15 & 0.70 \\ 1 & 0.25 & 0.40 & 0.35 \\ 2 & 0.10 & 0.65 & 0.25 \end{array}

How to compute $P(E(B-A)=X)$, that is the probability of $E$ being $X$?

1

There are 1 best solutions below

0
On BEST ANSWER

$\newcommand{\P}{\mathbb{P}}$From what I can see, $E$ is a random variable, with the rows of your table giving us the conditional distribution of $E$ given the value of $|B-A|$ (I am assuming it is $|B-A|$ because the table only has non-negative values but $B-A$ could he negative). So the first column of your table is telling us $\P(E = X\mid |B-A|)$. Hence to get $\P(E=X)$, you should use the law of total probability, which tells us that $$\P(E=X) = \sum_{k=0}^2 \color{blue}{\P(E=X\mid |B-A| =k)}\color{red}{\P(|B-A|=k)}.$$

The $\color{blue}{\P(E=X\mid |B-A|=k)}$ are in the first column of your table. So, compute the probabilities $\color{red}{\P(|B-A| = k)}$ for $k=0,1,2$ (make sure you know how to do this!) and you will have everything you need to work out your desired probability.