A probability generating function (PGF) is a generating function whose coefficients are the probabilities of a nonnegative integer-valued random variable. It represents an entire probability distribution by a single function. Its algebraic and analytic properties make it especially useful for studying counts, sums of independent random variables, random sums, and population models.
Definition and domain
Let take values in , with probability mass function
The probability generating function of is
where denotes expectation. At , the term corresponding to is defined to be , so that .
Although real arguments suffice for many probabilistic applications, the PGF can also be treated as a function of a complex variable. Its defining power series converges absolutely for every , because
Consequently, its radius of convergence is at least . It is an analytic function on the open unit disk and continuous on the closed unit disk.
The radius may be larger than , but this is not guaranteed. Finite-support distributions have polynomial PGFs, whereas sufficiently heavy-tailed distributions can have radius of convergence exactly .
A power series
is a PGF precisely when its coefficients are nonnegative and sum to .
Recovering probabilities
The coefficients of a PGF are exactly the probabilities it encodes. In particular,
where . Thus the PGF uniquely determines the distribution.
Using coefficient-extraction notation,
For example, if
then takes the values with probabilities .
Cumulative probabilities can also be extracted from a generating function. For ,
Similarly,
The second identity gives the tail-sum formula
with both sides allowed to be infinite.
Basic properties
Every PGF satisfies
On the interval , it is nondecreasing and convex. Within , all derivatives are nonnegative, since termwise differentiation gives
At selected complex arguments, a PGF can reveal arithmetic features of the distribution. For example,
so
More generally, values at roots of unity determine probabilities for residue classes. If , then
This is a finite Fourier inversion identity.
Factorial moments, mean, and variance
Differentiation at yields factorial moments. Define the falling factorial by
Then
The notation means the limit as increases to from below. These identities remain valid with value , even when the relevant moments are not finite.
In particular,
and
When the second moment is finite, the variance is
This follows from .
Ordinary moments can also be obtained with the differential operator
Since ,
A PGF therefore always exists on , even when the random variable has infinite mean or higher moments.
Standard examples
Bernoulli distribution
For a Bernoulli random variable with success probability ,
and
Binomial distribution
If has a binomial distribution with parameters and , then
The coefficients follow from the binomial theorem. Differentiation gives
Poisson distribution
For a Poisson distribution with parameter ,
Consequently,
Geometric distribution
Suppose counts failures before the first success in independent trials with success probability , where . Then
and the geometric series gives
If instead the variable counts trials up to and including the first success, its PGF is
The distinction reflects the two common conventions for the geometric distribution.
Independent sums and convolution
If and satisfy statistical independence, then
Multiplication of PGFs corresponds to convolution of probability mass functions:
For independent ,
If they are identically distributed with PGF , this becomes .
For example, independent Poisson variables with parameters have a sum with PGF
which identifies the sum as Poisson with parameter .
Without independence, the product rule generally fails.
Random sums and thinning
Random sums
Let be a nonnegative integer-valued random variable, and let be independent, identically distributed nonnegative integer-valued variables, independent of . Define
with when .
Conditioning on gives
Thus a random sum corresponds to composition of PGFs.
When the relevant moments are finite,
and
If is Poisson with parameter , the resulting compound Poisson distribution has PGF
Independent thinning
Suppose each of objects is retained independently with probability . If is the number retained, then its conditional distribution given is binomial:
Hence
In particular, thinning a Poisson count with mean produces a Poisson count with mean .
Branching processes
PGFs are central to branching processes. In a Galton–Watson process, each individual independently produces a random number of offspring with the same PGF . Let be the population in generation , starting from .
Conditional on , the next generation is a sum of independent offspring counts. Therefore,
and
where denotes -fold composition, not an ordinary power.
The probability of extinction by generation is
The eventual extinction probability is the smallest solution in of the fixed-point equation
Writing for the mean number of offspring:
- If , extinction occurs with probability .
- If , extinction occurs with probability , except when every individual has exactly one offspring.
- If , the extinction probability is less than .
Multivariate probability generating functions
For nonnegative integer-valued random variables , the joint PGF is
Its coefficients encode the joint distribution.
Setting all variables except equal to yields the marginal PGF of . Setting every argument equal to yields the PGF of , whether or not the variables are independent.
Mixed derivatives give mixed factorial moments. For distinct and , with finite second moments,
where boundary derivatives are interpreted as limits from within the domain.
The variables are mutually independent exactly when the joint PGF factors into their marginal PGFs.
Related transforms and limitations
The moment generating function
is related to the PGF by
where the expectation is finite. A PGF is always finite on , whereas need not be finite for any positive .
The characteristic function satisfies
and the Laplace transform satisfies
The ordinary PGF is tailored to nonnegative integer-valued variables. Variables with negative integer values generally require series with negative powers, while noninteger-valued variables are more naturally handled by other transforms.
Although a PGF determines every probability, extracting high-order coefficients can be computationally difficult. Repeated numerical differentiation may be unstable; depending on the representation, coefficient recurrences, symbolic expansion, or Fourier-based methods may be preferable. Analytic continuation beyond the original disk of convergence also requires care: a continued function value need not equal a finite expectation .