A discrete random variable has two possible values X = 1 or X = 0 with probabilities p and 1 − p, respectively. The probability density function of X is shown in the figure :
[Figure, described in words: two impulses on the x axis, of strength 1 − p at x = 0 and of strength p at x = 1.]
If Y = Σᵢ₌₁ⁿ Xᵢ, where Xᵢ, i = 1, 2, ..., n are statistically independent and identically distributed random variables, what is the probability distribution function of Y?
10 marks · directive: Derive
The core of a good answer
Y is a binomial random variable: P(Y = k) = ⁿC_k pᵏ(1 − p)ⁿ⁻ᵏ for k = 0, 1, ..., n, so its density is f_Y(y) = Σ ⁿC_k pᵏ(1 − p)ⁿ⁻ᵏ δ(y − k) and its distribution function is F_Y(y) = Σ_{k ≤ y} ⁿC_k pᵏ(1 − p)ⁿ⁻ᵏ, with mean np and variance np(1 − p).
An outline written by upsc.help, not an official UPSC answer; every fact checked by a separate review. How we work
What the question asks
The question asks for the distribution of a sum of n independent Bernoulli variables. It can be derived by counting arguments or more formally by convolution of densities, which is equivalent to multiplying characteristic or generating functions.
Answer outline
Introduction
Each Xᵢ is Bernoulli: f_X(x) = (1 − p)δ(x) + pδ(x − 1). Y counts how many of the n trials take the value 1, so Y takes integer values 0 to n.
Step 1: Generating function of one Xᵢ
- G_X(z) = E[zˣ] = (1 − p) + pz
- Equivalently the characteristic function φ_X(ω) = (1 − p) + p e^{jω}
Step 2: Sum of independent variables
- For independent Xᵢ the density of the sum is the n-fold convolution, so generating functions multiply
- G_Y(z) = [(1 − p) + pz]ⁿ
Step 3: Expand by the binomial theorem
- G_Y(z) = Σₖ₌₀ⁿ ⁿC_k pᵏ(1 − p)ⁿ⁻ᵏ zᵏ
- Coefficient of zᵏ: P(Y = k) = ⁿC_k pᵏ(1 − p)ⁿ⁻ᵏ, k = 0, 1, ..., n, with ⁿC_k = n!/[k!(n − k)!]
Step 4: Density and distribution function
- f_Y(y) = Σₖ₌₀ⁿ ⁿC_k pᵏ(1 − p)ⁿ⁻ᵏ δ(y − k)
- F_Y(y) = P(Y ≤ y) = Σₖ₌₀^{⌊y⌋} ⁿC_k pᵏ(1 − p)ⁿ⁻ᵏ for 0 ≤ y < n; F_Y(y) = 0 for y < 0 and 1 for y ≥ n (a staircase)
Step 5: Moments as a check
- E[Y] = G_Y′(1) = np; Var(Y) = np(1 − p)
- Σ P(Y = k) = (p + 1 − p)ⁿ = 1
Conclusion
Y is binomially distributed with parameters n and p. Its pdf is a set of n + 1 impulses at y = 0, 1, ..., n of strengths ⁿC_k pᵏ(1 − p)ⁿ⁻ᵏ, and its distribution function is the corresponding staircase.
Examples and data to cite
- In digital communication, the number of bit errors in an n-bit block over a binary symmetric channel with bit error probability p is binomial in exactly this way.
- For large n the binomial distribution approaches a Gaussian with mean np and variance np(1 − p) (De Moivre-Laplace theorem).
Facts to remember
- Bernoulli variable: mean p, variance p(1 − p).
- The pdf of a sum of independent random variables is the convolution of their pdfs.
- Binomial pmf: ⁿC_k pᵏ(1 − p)ⁿ⁻ᵏ; mean np; variance np(1 − p).
- A probability distribution function (CDF) of a discrete variable is a non-decreasing staircase from 0 to 1.