STAT0008 Statistical Inference: Level 6 and 7 Examination 2020 Page 1 STAT0008: STATISTICAL INFERENCE 2020 • Answer ALL questions. • You may submit only one answer to each question. • The relative weights attached to each question are as follows: Q1 (18), Q2 (26), Q3 (27), Q4 (29). • The numbers in square brackets indicate the relative weights attached to each part ques- tion. • Marks are awarded not only for a final answer but also for the clarity and coherence of your solution. • Information regarding some common probability distributions is provided at the begin- ning of the examination paper. Administrative details • This is an open-book examination. You may use your course materials to answer ques- tions. • You may not contact the course lecturer with any questions, even if you wish to clarify something or report an error on the paper. If you have any doubts concerning an examination question, please make a note in your answer to explain any assumptions that you have made. • UCL requires that all 24-hour online examinations have a specified overall word limit. The overall word limit for this examination has been set well in excess of the expected amount of work so that you need not worry about exceeding it. Therefore, we expect that solutions to the paper will be much shorter than the specified word limit. Turn Over STAT0008 Statistical Inference: Level 6 and 7 Examination 2020 Page 2 Formatting your solutions for submission • For all questions, you may choose to type or hand-write your answers. • You should submit ONE document that contains your solutions for all questions. Please follow the UCL guidance on combining text and photographed/scanned work, if applica- ble. • Please ensure that your handwritten solutions are clear and are readable in the document that you submit. You are encouraged to write out solutions neatly, prior to submission, once you are happy with them. Plagiarism and collusion • You must work alone. In particular, any discussion of the paper with anyone else is not acceptable. You are encouraged to read the Department of Statistical Science’s advice on collusion and plagiarism, which you can find here. • Parts of your submission will be screened via Turnitin to check for plagiarism and collu- sion. • If there is any doubt as to whether the solutions you submit are entirely your own work you may be required to participate in an investigatory viva to establish authorship. Continued STAT0008 Statistical Inference: Level 6 and 7 Examination 2020 Page 3 Common Probability Distributions Binomial Distribution Notation: X ∼ Bin(n, p) Probability mass function: P(X = x) = ( n x ) px(1− p)n−x with p ∈ [0, 1], n ∈ N Support: x ∈ {0, 1, . . . , n} Mean: E(X) = np Variance: Var(X) = np(1− p) Poisson Distribution Notation: X ∼ Poi(λ) Probability mass function: P(X = x) = λxe−λ x! with λ ∈ [0,∞) Support: x ∈ {0, 1, . . .} Mean: E(X) = λ Variance: Var(X) = λ Negative Binomial Distribution Notation: X ∼ NegBin(k, p) Probability mass function: P(X = x) = ( x−1 k−1 ) pk(1− p)x−k with p ∈ [0, 1], k ∈ N Support: x ∈ {k, k + 1, . . .} Mean: E(X) = k p Variance: Var(X) = k(1−p) p2 Turn Over STAT0008 Statistical Inference: Level 6 and 7 Examination 2020 Page 4 (Continuous) Uniform Distribution Notation: X ∼ U [a, b] Probability density function: fX(x; a, b) = 1 b−a with (a, b) ∈ R2, a < b Support: x ∈ [a, b] Mean: E(X) = 1 2 (a+ b) Variance: Var(X) = 1 12 (b− a)2 Normal Distribution Notation: X ∼ N (µ, σ2) Probability density function: fX(x;µ, σ 2) = 1√ 2piσ2 exp {− 1 2σ2 (x− µ)2} with µ ∈ R, σ ∈ (0,∞) Support: x ∈ (−∞,∞) Mean: E(X) = µ Variance: Var(X) = σ2 Exponential Distribution Notation: X ∼ Exp(λ) Probability density function: fX(x;λ) = λe −λx with λ ∈ (0,∞) Support: x ∈ (0,∞) Mean: E(X) = 1 λ Variance: Var(X) = 1 λ2 Gamma Distribution Notation: X ∼ Gamma(α, β) Probability density function: fX(x;α, β) = βα Γ(α) xα−1e−βx with α ∈ (0,∞), β ∈ (0,∞) Support: x ∈ (0,∞) Mean: E(X) = α β Variance: Var(X) = α β2 Continued STAT0008 Statistical Inference: Level 6 and 7 Examination 2020 Page 5 Q1. Suppose that X1, . . . , Xn are independent and identically distributed Poisson random variables, such that Xi ∼ Poi(µ) where µ > 0 and i ∈ {1, . . . , n}. (a) Determine the maximum likelihood estimator of µ, denoting this estimator µˆ. [3] (b) Determine the asymptotic (large sample) distribution of µˆ. [4] (c) Is µˆ weakly consistent for µ? Justify your answer. [3] Suppose that Y1, . . . , Yn are independent and identically distributed exponential random variables, such that Yi ∼ Exp(λ) where λ > 0 and i ∈ {1, . . . , n}. The random variable Z is defined as the minimum of {Y1, . . . , Yn}. (d) Determine P(Z > z) for z > 0 and hence show that λZ is a pivotal quantity for λ. [5] (e) Using λZ, construct a two-sided, equal-tailed, 95% confidence interval for λ, based on the sample Y1, . . . , Yn. [3] Turn Over STAT0008 Statistical Inference: Level 6 and 7 Examination 2020 Page 6 Q2. The random variables X1, . . . , Xn are independent and sampled from a distribution with probability density function f(x;ψ) = 1 ψ exp ( 1 ψ − 1 ψ ex + x ) x ≥ 0; 0 otherwise; where ψ > 0 is an unknown parameter. (a) Determine the maximum likelihood estimator of ψ and call this estimator ψˆ. [3] (b) Is ψˆ is the minimum variance bound unbiased estimator (MVBUE) of ψ? Justify your answer. [3] (c) Determine the Crame´r-Rao lower bound for the variance of unbiased estimators of 1/ψ. [3] (d) Determine the median of the distribution of Xi, for any i ∈ {1, . . . , n}. [6] A sample of X1, . . . , Xn, denoted x1, . . . , xn, is taken where n = 60, ∑60 i=1 xi = 45.95 and∑60 i=1 e xi = 219.73. (e) Construct an approximate 95% confidence interval for ψ. You may find the following quantiles of the standard N (0, 1) distribution helpful: z0.025 = 1.96, z0.05 = 1.645, z0.1 = 1.28 where, for a given value ∈ (0, 1), z is the quantile such that P(Z > z) = for Z ∼ N (0, 1). [4] A Bayesian prior distribution for ψ is defined as pi(ψ) = βα Γ(α) 1 ψα+1 exp ( −β ψ ) . where α > 0 and β > 0 are known parameters. (f) Derive the posterior distribution of ψ and hence determine the Bayes’ estimate of ψ under quadratic error loss. [7] Continued STAT0008 Statistical Inference: Level 6 and 7 Examination 2020 Page 7 Q3. Suppose that X1, . . . , Xn are independent and identically distributed Bin(1, p) random variables. The parameter θ is such that θ = p2. (a) Find a minimal sufficient statistic for θ. [4] (b) Determine the maximum likelihood estimator of θ. Is this estimator unbiased for θ? Justify your answer. [7] The estimator U is defined U = { 1 if X1 = 1 and X2 = 1 0 otherwise (c) Show that U is unbiased for θ. [1] (d) Using the Rao-Blackwell theorem, determine an unbiased estimator of θ, denoting this estimator θ˜, such that Var(θ˜) ≤ Var(U). [7] (e) Is θ˜ the minimum variance unbiased estimator (MVUE) of θ? Justify your answer. [4] (f) Does the variance of θ˜ attain the Crame´r-Rao lower bound for the variance of unbi- ased estimators of θ? Justify your answer. [4] Turn Over STAT0008 Statistical Inference: Level 6 and 7 Examination 2020 Page 8 Q4. The random variables X1, . . . , Xn are independent and identically distributed, where Xi ∼ Exp(λ) for i ∈ {1, . . . , n} and λ > 0. (a) Based on the sample, X1, . . . Xn, construct the most powerful test of size α of the hypotheses H0 : λ = λ0 against H1 : λ = λ1 where α ∈ (0, 1), λ0 > 0, λ1 > 0 and λ1 > λ0. You may use the following results without proof: If Z1, . . . , Zn are independent Exp(θ) random variables then ∑n i=1 Zi ∼ Gamma(n, θ). If n ∈ N and Y ∼ Gamma(n, θ) then 2θY ∼ χ22n. If U ∼ χ2ν then P(U > χ2ν()) = for ∈ (0, 1). [9] (b) Based on the sample X1, . . . , Xn, construct a 100(1 − α)% confidence lower bound for λ. [3] (c) Derive the power function for the test that you constructed in (a). Your answer may be written in terms of a cumulative distribution function, F (.), where the explicit form of F need not be evaluated. [3] (d) Is the test that you constructed in (a) uniformly most powerful and of size α for a test of the hypotheses H0 : λ ≤ λ0 against H1 : λ > λ0? Justify your answer. [4] Associated with X1, . . . , Xn is the set of random variables {Z1, . . . , Zn} such that Zi = { 0 if Xi < γ; 1 if Xi ≥ γ; for i ∈ {1, . . . , n}, where γ > 0 is a known parameter. (e) Write down the likelihood function, L(λ, z), based on an observed sample z = (z1, . . . , zn) > of Z1, . . . , Zn and hence determine the maximum likelihood estimate of λ. [4] (f) State a condition that must be satisfied by X1, . . . , Xn to ensure that the maximum likelihood estimate of λ, derived in (e), is finite. [2] (g) Determine the Crame´r-Rao lower bound for the variance of unbiased estimators of λ, where λ is to be estimated using z1, . . . , zn. [4] End of Paper
欢迎咨询51作业君