Conditional probability based on assuming linear relationships

38 Views Asked by At

I am a statistical novice trying to work something out which is probably basic. Or impossible. I have discrete probability distributions for two random variables $A$ and $B$, something like this:

k P(A=k) P(B=k)
0 0 0.1
1 0 0.4
2 0.1 0.4
3 0.15 0.1
4 0.25 0
5 0.25 0
6 0.15 0
7 0.1 0
8 0 0
9 0 0

I want to find probabilities like $P(A=k \cap B=j)$. When $A$ and $B$ are independent, I can just multiply $P(A=k)$ and $P(B=j)$. But in my setting I want to try to model situations where $A$ and $B$ are not independent.

As a very simple case, I'd like to assume that maybe there is some linear relationship between A and B, so that, say, if A ends up observed higher than expected, then B will probably also end up higher than expected. This linear relationship would have a constant built in, say $C$. I'm imagining $C$ is between -1 and 1, where 1 means they are directly correlated, 0 means they are independent, and -1 means inversely correlated.

I assume this kind of thing is well-known- what is the appropriate language and framework for me to be using? Ideally I'd like a formula for computing $P(A=k \cap B=j)$ in terms of $C$ and $P(A=k)$ and $P(B=j)$.