STATISTICAL THEORY FOR CS EXAM 2017 A PAVEL CHIGANSKY Problem 1. Let Z1 , Z2 , ... be i.i.d. Ber(θ), θ ∈ (0, 1) r.v.’s. We have seen that there is no unbiased estimate of the odds θ/(1 − θ), based on Z1 , ..., Zn for any fixed n ≥ 1. Let X1 = min{n : Zn = 0}, be the number of tosses till the first 0 occurs. As is well known, X1 has Geo(1 − θ) distribution1: Pθ (X1 = k) = θk−1 (1 − θ), k ∈ {1, 2, ...}. (1) Find the MLE of θ on the basis of X1 . (2) Find the MLE of η(θ) := θ/(1 − θ). Hint: η(θ) is a one-to-one function of θ ∈ (0, 1). (3) Show that MLE η̂ is unbiased. Explain the “contradiction” with the non-existence claim above. (4) Is the obtained estimate above UMVUE ? Calculate its MSE risk. (5) Calculate the C-R lower bound for the MSE risk of unbiased estimates of θ/(1 − θ). Is C-R bound attained ? (6) Encouraged by his progress, the statistician suggests to count the tosses till m ≥ 1 zeros occur: n n X o Xm = min n : (1 − Zi ) = m . i=1 Argue that the p.m.f. of Xm is given by: k−1 pXm (k; θ) = Pθ (Xm = k) = θk−m (1 − θ)m , m−1 k≥m (7) Find the MLE of η(θ) = θ/(1 − θ) on the basis of Xm . (8) Do MLEs from the previous question form a consistent sequence of estimates of θ/(1 − θ) ? P Hint: Show that Xm = m j=1 ξi , where ξi are i.i.d. Geo(1 − θ) r.v.’s and apply the LLN. (9) Find the asymptotic rate and the asymptotic error distribution of the MLE’s found above 1E = θ 1 1−θ and varθ (X1 ) = θ (1−θ)2 1 2 PAVEL CHIGANSKY Problem 2.(Change Detection) A plant produces bottles of mineral water. Production is monitored by a statistician, who samples the water acidity daily. She suspects that the production line stopped to work properly within the last n days and suggests to perform a test. Let (X1 , ..., Xn ) denote the acidity of the water measured on each one of the days and let θ ∈ {1, ..., n} denote the day index at which the change have occurred. The statistician wants to test the hypothesis: H0 : X1 , ..., Xn are i.i.d. N (0, 1) r.v.’s against the alternatives ( X1 , ..., Xθ−1 ∼ N (0, 1) H1 : X1 , ..., Xn are independent, Xθ , ..., Xn ∼ N (a, 1) θ ∈ {1, ..., n} where a > 0 is a known constant. (1) Find the likelihood function of the data under the alternative (2) For n = 1, find the most powerful level α test (3) Find the α level test, using the test statistic X̄n and calculate its power function. (4) Prove that the GLRT test with a critical value c rejects H0 iff n X max (Xi − a/2) > c. θ∈{1,...,n} i=θ (5) Find the GLRT statistic, if a is unknown and is to be considered as a parameter as well. Department of Statistics, The Hebrew University, Mount Scopus, Jerusalem 91905, Israel E-mail address: pchiga@mscc.huji.ac.il