Define multivariate normal distribution and write down its important properties. Let X₁, X₂, X₃ and X₄ are mutually independent random vectors following Nₚ(μ, Σ). Then find the marginal distributions of
(i) W₁ = (1/4)X₁ − (1/4)X₂ + (1/4)X₃ − (1/4)X₄
(ii) W₂ = (1/4)X₁ + (1/4)X₂ − (1/4)X₃ − (1/4)X₄
Also obtain the joint distribution of random vectors W₁ and W₂.
20 marks · directive: Define
The core of a good answer
W₁ and W₂ are each Nₚ(0, Σ/4), because the coefficients sum to 0 (so the means cancel) and their squares sum to 4/16; since ∑c₁ᵢc₂ᵢ = (1 − 1 − 1 + 1)/16 = 0, Cov(W₁, W₂) = 0, so (W₁′, W₂′)′ ~ N₂ₚ((0, 0)′, [Σ/4 0; 0 Σ/4]) and W₁, W₂ are independent.
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
Define the distribution (density form for non-singular Σ, or the 'every linear combination is normal' definition), list its key properties, and then apply the linear-combination property: ∑cᵢXᵢ ~ Nₚ(∑cᵢμ, ∑cᵢ²Σ) for independent Xᵢ, with Cov(∑c₁ᵢXᵢ, ∑c₂ᵢXᵢ) = (∑c₁ᵢc₂ᵢ)Σ.
Answer outline
Introduction
X ~ Nₚ(μ, Σ), Σ positive definite, if its density is f(x) = (2π)^{−p/2}|Σ|^{−1/2} exp[−½(x − μ)′Σ⁻¹(x − μ)], x ∈ Rᵖ. Equivalently, a′X is univariate normal for every a.
Important properties
- Characteristic function exp(it′μ − ½t′Σt); E(X) = μ, D(X) = Σ.
- AX + b ~ N_q(Aμ + b, AΣA′) for any q × p matrix A.
- All marginal distributions are multivariate normal.
- Uncorrelated components (zero covariance blocks) are independent.
- Conditional distributions are normal: X₁ | X₂ ~ N(μ₁ + Σ₁₂Σ₂₂⁻¹(x₂ − μ₂), Σ₁₁ − Σ₁₂Σ₂₂⁻¹Σ₂₁).
- (X − μ)′Σ⁻¹(X − μ) ~ χ²ₚ; contours of the density are ellipsoids.
Step 1: Linear combinations of independent vectors
- If X₁, …, X₄ are independent Nₚ(μ, Σ), then ∑cᵢXᵢ ~ Nₚ((∑cᵢ)μ, (∑cᵢ²)Σ).
Step 2: W₁
- c = (1/4, −1/4, 1/4, −1/4): ∑cᵢ = 0, ∑cᵢ² = 4/16 = 1/4.
- W₁ ~ Nₚ(0, Σ/4).
Step 3: W₂
- d = (1/4, 1/4, −1/4, −1/4): ∑dᵢ = 0, ∑dᵢ² = 1/4.
- W₂ ~ Nₚ(0, Σ/4).
Step 4: Covariance
- Cov(W₁, W₂) = ∑cᵢdᵢ Σ = (1/16)(1 − 1 − 1 + 1)Σ = 0.
Step 5: Joint distribution
- (W₁′, W₂′)′ is a linear transformation of (X₁′, …, X₄′)′, hence 2p-variate normal:
- (W₁′, W₂′)′ ~ N₂ₚ(0, [Σ/4 0; 0 Σ/4]).
- Zero covariance under joint normality ⇒ W₁ and W₂ are independent.
Conclusion
Both W₁ and W₂ are Nₚ(0, ¼Σ), and they are independent with joint distribution N₂ₚ(0, diag(Σ/4, Σ/4)). The coefficient vectors are orthogonal contrasts, which is why the means vanish and the vectors are uncorrelated.
Examples and data to cite
- Check: W₁ + W₂ = (X₁ − X₄)/2 ~ Nₚ(0, Σ/2), which equals Σ/4 + Σ/4 as required for independent W₁ and W₂.
Facts to remember
- Density of Nₚ(μ, Σ): (2π)^{−p/2}|Σ|^{−1/2}exp[−½(x − μ)′Σ⁻¹(x − μ)].
- Linear transformation: AX + b ~ N(Aμ + b, AΣA′).
- For jointly normal vectors, zero covariance ⇔ independence.
- ∑cᵢXᵢ ~ Nₚ(∑cᵢμ, ∑cᵢ²Σ) for independent Xᵢ ~ Nₚ(μ, Σ).