We have seen that the marginal distribution is a distribution that describes a subset of random variables. Next, we will discuss a distribution that describes all the random variables in the set. This is called a joint distribution. Let us look at the joint distribution that involves the Degree score and Experience random variables in the job hunt example:
Degree score |
Relevant Experience | |||
Highly relevant |
Not relevant | |||
Poor |
0.1 |
0.1 |
0.2 | |
Average |
0.1 |
0.4 |
0.5 | |
Excellent |
0.2 |
0.1 |
0.3 | |
0.4 |
0.6 |
1 |
The values within the dark gray-colored cells are of the joint distribution, and the values in the light gray-colored cells are of the marginal distribution (sometimes called that because they are written on the margins). It can be observed that the individual marginal distributions sum up to 1, just like a normal probability distribution.
Once the joint distribution is described, the marginal distribution can be found by summing up individual rows or columns. In the preceding table, if we sum up the columns, the first column gives us the probability for Highly relevant, and the second column for Not relevant. It can be seen that a similar tactic applied on the rows gives us the probabilities for degree scores.