Let X₁, X₂, ⋯, X_N be a random sample from Nₚ(μ, Σ). Show that Hotelling’s T²-statistic for testing H₀ : μ = μ₀ against H₁ : μ ≠ μ₀ is a generalization of t-statistic for testing H₀ : μ = μ₀ against H₁ : μ ≠ μ₀ in the univariate case. Calculate T²-statistic for testing H₀ : μ = (9, 5)′ on the basis of the data taken from a bivariate normal distribution given below :
X = [6 9; 10 6; 8 3] (rows (6, 9), (10, 6), (8, 3))
15 marks · directive: Show
The core of a good answer
T² = N(x̄ − μ₀)′S⁻¹(x̄ − μ₀) reduces for p = 1 to N(x̄ − μ₀)²/s² = t²; for the data, x̄ = (8, 6)′, S = [4 −3; −3 9], S⁻¹ = (1/27)[9 3; 3 4], and T² = 3 × 7/27 = 7/9 ≈ 0.778, so F = T²/4 = 0.194 on (2, 1) d.f., far below 199.5, and H₀ is not rejected at 5%.
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
Show the reduction from T² to t² (and the matching F distribution), then compute the sample mean vector, the unbiased covariance matrix S (divisor N − 1), its inverse and the quadratic form. Finish with the F conversion and decision, using the F table supplied with the paper.
Answer outline
Introduction
T² = N(x̄ − μ₀)′S⁻¹(x̄ − μ₀), S = ∑(Xᵢ − x̄)(Xᵢ − x̄)′/(N − 1). Under H₀, [(N − p)/(p(N − 1))]T² ~ F(p, N − p). Here N = 3, p = 2.
Step 1: Reduction to t
- For p = 1: S = s², so T² = N(x̄ − μ₀)²/s² = [(x̄ − μ₀)/(s/√N)]² = t².
- The F relation gives F(1, N − 1) = t²ₙ₋₁; rejecting for large T² is the two-sided t-test.
Step 2: Mean vector
- x̄ = ((6 + 10 + 8)/3, (9 + 6 + 3)/3)′ = (8, 6)′; x̄ − μ₀ = (−1, 1)′.
Step 3: Covariance matrix
- Deviations: (−2, 3), (2, 0), (0, −3).
- s₁₁ = (4 + 4 + 0)/2 = 4; s₂₂ = (9 + 0 + 9)/2 = 9; s₁₂ = (−6 + 0 + 0)/2 = −3.
- S = [4 −3; −3 9]; |S| = 36 − 9 = 27; S⁻¹ = (1/27)[9 3; 3 4].
Step 4: T²
- (x̄ − μ₀)′S⁻¹(x̄ − μ₀) = (1/27)[9(1) + 2 × 3 × (−1)(1) + 4(1)] = 7/27.
- T² = 3 × 7/27 = 7/9 = 0.778.
Step 5: Test
- F = (N − p)T²/[p(N − 1)] = (1/4)(7/9) = 0.194 with (2, 1) d.f.
- F₀.₀₅(2, 1) = 199.5 (table); 0.194 < 199.5, so H₀ : μ = (9, 5)′ is not rejected.
Conclusion
T² = 0.778 and the data are consistent with μ = (9, 5)′. With only N = 3 observations in 2 dimensions the test has a single denominator degree of freedom and almost no power. Values computed in Python.
Examples and data to cite
- Harold Hotelling proposed the T² distribution in 1931 as a generalisation of Student's t-distribution.
Facts to remember
- Hotelling's T² = N(x̄ − μ₀)′S⁻¹(x̄ − μ₀).
- [(N − p)/(p(N − 1))]T² ~ F(p, N − p) under H₀.
- For p = 1, T² = t² and F(1, ν) = t²_ν.