P n=0 z pn for the following distributions, and hence calculate the first and second cumulants. n (a) pn = e−α αn! α > 0, n ≥ 0 (Poisson distribution)
N (b) pn = n αn (1−α)N −n 0 < α < 1, 0 ≤ n ≤ N (binomial distribution)
In the second case, the sum in the generating function terminates at
N. Additional exercise: By treating n as a continuous variable, de- termine its most probable value in (a). You may also use Stirling’s approximation to simplify n!.
2. Consider the Cauchy-Lorentz (also known as Lorentz) distribution
a 1 P (x) = π (x + a2 ) 2
defined in the range −∞ < x < ∞.
(i) Show, by direct integration none of its moments (except the zero’th one, of course) exist. (ii) Compute the characteristic (generating) function G(k) for k < 0 and k > 0 separately. Why is this separation necessary? Discuss the behaviour of the function near k = 0 and connect with the property discussed earlier.
3. Consider the probability distribution P (x) = λe−λx defined for x ∈
[0, ∞), while P (x) = 0 for x < 0 (here, λ > 0). Find its cumulant generator and hence the first two cumulants.
4. The geometric distribution pn = (1 − γ)γ n is defined for n ≥ 0, and
γ < 1. Derive a general expression for its m’th cumulant, where m ≥ 1.
5. Consider P (x) = n pn δ(x − xn ) with n pn = 1 and n denote a set of
P P
integers. Derive a general expression for the variance of x.