1. Joint Probability Mass and Density Functions

Definition — Joint pmf (discrete)

is a joint pmf of discrete RVs if:

  1. for all
  2. , and for a region :

Definition — Joint pdf (continuous)

is a joint density of continuous RVs if:

  1. (volume under the surface over )

Worked Example — Discrete Joint Table

= errors found in module A, = errors found in module B of a program:

0.100.200.100.40
0.150.250.050.45
0.050.050.050.15
0.300.500.201.00

Worked Example — Joint Density


Valid: ✓.

: the region is the triangle under the line .

Region X + Y < 1 in the unit square y = 1 − x P = 1/3 0 0 1 1 x y

⚠️ Always sketch the region first; the limits of the inner integral depend on the outer variable.


2. Marginal Distributions

Definition — Marginal Distributions


The marginal of is just the distribution of alone; in a table it is the row/column total (hence “marginal” — written in the margin).

Worked Example

  • Table above: and .
  • Density : for ; by symmetry .

3. Conditional Distributions

Definition — Conditional Distribution


For fixed , is a genuine pmf/pdf in (it sums/integrates to 1).

Worked Example — Discrete

Distribution of given ():

012

Worked Example — Continuous

For :

⚠️ For continuous , , so "" can’t use the event definition ; it is defined through the conditional density.


4. Statistical Independence

Definition — Independent RVs

and are statistically independent iff

Equivalently, : knowing tells nothing about .

Worked Example — Checking Independence

  • Table: but → dependent (one failing cell is enough).
  • : → dependent.
  • on the unit square: , , and → independent.

Theorem — Factorization Shortcut

If the support is a rectangle and factors into a function of times a function of , then and are independent.

⚠️ The support matters: on looks like it factors, but the triangular support links and , so , are dependent.

Definition — Several RVs

are mutually independent iff

Then, e.g., .

Example. Three independent components have lifetimes with . Probability all three last beyond 1000 h: .