Conditional Distribution

The conditional distribution is a fundamental idea in probability and statistics — it describes how one random variable behaves once we know something about another. It sits at the heart of regression, prediction and Bayesian reasoning. This guide explains what a conditional distribution is, how it relates to joint and marginal distributions, the formulas, and why it matters — in clear, plain language. It complements our guide to the best linear unbiased estimator and is relevant to anyone studying statistics, econometrics or data analysis.

What is a conditional distribution?

A conditional distribution is the probability distribution of one random variable given that another variable takes a particular value. We write it as the distribution of Y given X = x. The idea is simple but powerful: knowing the value of X often changes what we expect of Y. For example, the distribution of someone's weight is one thing in general, but the distribution of weight given a particular height is narrower and shifted. The conditional distribution captures exactly this — how the behaviour of Y is updated once we condition on information about X.

Joint, marginal and conditional distributions

To place the conditional distribution in context, it helps to distinguish three related ideas:

• The joint distribution describes the probabilities of X and Y together — how they co-vary.
• The marginal distribution describes one variable on its own, ignoring the other (the "unconditional" distribution).
• The conditional distribution describes one variable given a specific value of the other.

These three are tightly linked: the conditional distribution is what you get when you take a "slice" of the joint distribution at a fixed value of X and then rescale it so the probabilities add up to one.

The formulas

For discrete variables, the conditional probability of Y given X is:

P(Y = y | X = x) = P(X = x, Y = y) ÷ P(X = x)

That is, the joint probability divided by the marginal probability of the condition. For continuous variables, the same logic applies to density functions:

f(y | x) = f(x, y) ÷ f(x)

where f(x, y) is the joint density and f(x) is the marginal density of X. In both cases, dividing by the probability (or density) of the condition is what rescales the slice into a proper distribution.

A simple example

Suppose we roll two fair dice and let X be the first die and Y be the total of both. Unconditionally, the total Y ranges from 2 to 12 with the familiar triangular distribution. But given X = 6 (the first die shows a six), the total Y can only be 7 to 12, each with equal probability one-sixth — a completely different, flat distribution. Conditioning on X = 6 has reshaped what we expect of Y. This is the essence of a conditional distribution: new information about one variable rewrites the probabilities of the other.

Conditional mean and variance

Two summaries of a conditional distribution are especially useful. The conditional mean, E(Y | X = x), is the average value of Y for a given X — and as a function of x, it's exactly the regression function. This is the deep link between conditional distributions and regression: regression is, at heart, an attempt to estimate the conditional mean of Y given the predictors. The conditional variance, meanwhile, measures how spread out Y is for a given X. When the conditional variance is constant across values of x, we have homoskedasticity; when it changes, we have heteroskedasticity — both key concepts in regression analysis.

Independence and why it matters

A special case ties everything together: if X and Y are independent, then conditioning on X tells you nothing about Y, so the conditional distribution equals the marginal distribution. In other words, P(Y | X) = P(Y). When that's not the case, X carries information about Y — which is precisely what makes prediction possible. Conditional distributions are therefore central to regression and prediction (modelling Y given predictors), to Bayesian inference (updating beliefs given evidence via conditional probabilities), and to understanding any relationship between variables. They are the formal language for "what we expect, given what we know".

Frequently asked questions

What is a conditional distribution?

The probability distribution of one random variable given that another takes a particular value — how Y behaves once we condition on knowing X = x.

How does it relate to joint and marginal distributions?

It's a rescaled slice of the joint distribution at a fixed value of the conditioning variable; the marginal distribution ignores the other variable entirely.

What are the conditional distribution formulas?

Discrete: P(Y=y|X=x) = P(X=x,Y=y) ÷ P(X=x). Continuous: f(y|x) = f(x,y) ÷ f(x) — joint divided by the marginal of the condition.

How does it connect to regression?

The conditional mean E(Y|X=x), seen as a function of x, is the regression function — regression estimates the conditional mean of Y given the predictors.

Master statistics with Learnsignal

Concepts like conditional distributions underpin data analysis. Learnsignal's tutor-led ACCA and CIMA courses build the quantitative foundations — with flexible, supported online study that fits around work.