 Solution

ID: past-exam-of-the-mathematics-course-of-the-university-of-cambridge/2022/ib/paper-3/18h/c/solution

Codex  0 Created 2026-09-24 Updated 2026-09-25

Write

β = A Y

, where

A = (X^{T} Σ_{0}^{- 1} X)^{- 1} X^{T} Σ_{0}^{- 1}, A X = I .

Any other linear unbiased estimator is

C Y

with

CX = I

, so

C = A + D

and

D X = 0

. The cross covariance vanishes because

A Σ_{0} D^{T} = (X^{T} Σ_{0}^{- 1} X)^{- 1} X^{T} D^{T} = 0.

Therefore

Cov (C Y) - Cov (A Y) = σ^{2} D Σ_{0} D^{T} ⪰ 0.

β is the best linear unbiased estimator of β .

 New to topics? Read the docs here!