IAI Actuarial Core Principles · Actuarial Statistics

Jointly Distributed Random Variables: Joint, Marginal and Conditional

Jointly distributed random variables describe two or more random variables together through one joint pmf or pdf. From it you find marginals by summing or integrating, conditionals by dividing, and covariance, correlation and distributions of sums. Solve problems by fixing the support first, then integrating or summing in the right order.

What this chapter covers

This chapter moves you from one random variable to several. You describe X and Y together with a joint pmf p(x, y) or a joint pdf f(x, y). Probabilities come from summing or double integrating over a region. Most mistakes start with the region, so you learn to sketch the support early.

From the joint distribution you pull out the pieces. The marginal of X comes from summing or integrating out Y. The conditional distribution is f(x | y) = f(x, y) ÷ f_Y(y), for values of y where f_Y(y) > 0. You then test independence, and measure dependence with covariance and correlation. The chapter ends with the distribution of functions of random variables, such as sums and general transformations.

The ideas are used across the Actuarial Statistics module. In CS1 they underpin sampling distributions, estimators, regression and Bayesian statistics, where a joint distribution of data and parameter is central. In CS2 they support risk modelling, stochastic processes, time series and survival models. Treat this chapter as a base for both papers, not as a stand-alone topic.

Joint distributions are a base skill that later chapters assume. In the written section, questions often ask for a marginal, a conditional expectation, a covariance or the distribution of a sum, and each step carries marks for method. In the multiple-choice section, quick checks on independence, variance of a sum or correlation can be won in under two minutes if your formulas are secure. The same tools appear again in regression, Bayesian posterior work and risk models, so time spent here pays back in several other chapters. Paper B computer-based work also uses simulation and covariance calculations, so the concepts help there too.

Jointly distributed random variables: topics in the order to study them

1Joint Distributions of Discrete and Continuous VariablesEverything else is built from the joint pmf or pdf, and you must be comfortable with supports and double sums or integrals first.
2Marginal Distributions and IndependenceMarginals are the first thing you extract from a joint distribution, and the independence test uses them directly.
3Conditional Distributions and Conditional ExpectationConditionals need the joint and the marginal together, and they lead to E[X] = E[E[X | Y]] and the variance decomposition.
4Covariance and CorrelationOnce you can compute joint expectations such as E[XY], you can measure dependence and handle variances of sums.
5Transformations and Sums of Random VariablesThis topic uses all the earlier ones, so it comes last: joint supports, independence, moments and change of variable.

How to prepare Jointly distributed random variables

Aim to be fluent in a small set of formulas and a routine for setting up regions. Practise on paper, because the exam is written and you must show working.

Write the core definitions from memory: joint pmf and pdf, marginals, conditionals, independence, E[g(X, Y)], covariance and correlation.
For every continuous example, sketch the support before integrating, and mark the limits for both orders of integration.
Practise getting marginals and conditionals from the same joint distribution, then check that each one sums or integrates to 1.
Learn the tower rule and the variance rule: E[X] = E[E[X | Y]] and Var(X) = E[Var(X | Y)] + Var(E[X | Y]). Use them on a few mixture problems.
Do covariance questions with the shortcut Cov(X, Y) = E[XY] − E[X]E[Y], and practise Var(aX + bY) = a²Var(X) + b²Var(Y) + 2ab Cov(X, Y).
For transformations, practise the sum of independent variables using moment generating functions, and the change of variable method with the Jacobian for continuous cases.
Finish with timed past-paper style questions, then redo the ones where you lost marks without looking at your earlier answer.

Common mistakes in Jointly distributed random variables

Using wrong integration limits because the support was not sketched.
Fix: Always draw the support first. Write the limits for the inner variable in terms of the outer variable, and check the result integrates to 1.
Concluding independence just because the joint density factorises in form.
Fix: Check both the factorisation and that the support is a product of ranges for x and y.
Saying zero covariance means independence.
Fix: Remember that independence implies zero covariance only. Use a counterexample such as Y = X² with X symmetric about 0 to fix it in memory.
Forgetting the covariance term in the variance of a sum or difference.
Fix: Write the full formula first, then drop the covariance term only after stating that the variables are independent. For differences, the term is −2Cov(X, Y).
Treating E[X | Y] as a number instead of a random variable.
Fix: Keep E[X | Y] as a function of Y when using the tower rule or the variance decomposition, and take the outer expectation or variance over Y.
Omitting the Jacobian or the new support in a transformation.
Fix: Follow a fixed routine: invert the transformation, find the Jacobian, take its absolute value, then state the new range and check the density integrates to 1.

Last-day revision: Jointly distributed random variables

Joint pmf: p(x, y) ≥ 0 and Σ Σ p(x, y) = 1. Joint pdf: f(x, y) ≥ 0 and the double integral over the support equals 1.
Marginal of X: sum p(x, y) over y, or integrate f(x, y) over y, across the full range of y for that x.
Conditional: f(x | y) = f(x, y) ÷ f_Y(y), valid only where f_Y(y) > 0.
X and Y are independent if and only if f(x, y) = f_X(x) f_Y(y) for all x, y. The support must also be a rectangle.
E[X] = E[E[X | Y]] and Var(X) = E[Var(X | Y)] + Var(E[X | Y]).
Cov(X, Y) = E[XY] − E[X]E[Y].
Correlation ρ = Cov(X, Y) ÷ (σ_X σ_Y), and it always lies between −1 and 1.
Independent implies Cov = 0, but Cov = 0 does not imply independent.
Var(X + Y) = Var(X) + Var(Y) + 2Cov(X, Y). For independent variables the covariance term is zero.
Sum of independent variables: the MGF of X + Y is the product of the MGFs.
Sum of independent Poisson variables with means λ₁ and λ₂ is Poisson with mean λ₁ + λ₂.
Change of variable: multiply by the absolute value of the Jacobian, and recompute the new support.

Jointly distributed random variables practice questions

Jointly distributed random variables in other exams

The same ground in other exams, if you are preparing for more than one or want another angle on it.

Jointly distributed random variables: frequently asked questions

Is this chapter more important for CS1 or CS2?

Both use it. In CS1 it supports sampling distributions, estimation, regression and Bayesian statistics. In CS2 it supports risk modelling, stochastic processes and time series, so a strong grasp helps across the whole module.

Do I need to learn the Jacobian method for transformations?

Yes, you should be able to use it for continuous variables. Also learn the MGF method for sums of independent variables, as it is often faster and avoids the need for a change of variable.

How do I know if two random variables are independent?

Check that the joint pmf or pdf equals the product of the marginals for every pair of values. For continuous variables, also check that the support is a product of ranges. If either check fails, the variables are dependent.

Can correlation be used to prove independence?

No. Correlation only measures linear dependence. Two variables can have zero correlation and still be strongly dependent, so you need the full joint distribution to prove independence.