14.381: Statistics Fall, 2007 Midterm Examination 1. True, false, or uncertain? Explain your answer briefly. (a) If θˆ1 and θˆ2 are unbiased estimators for parameter θ, then 31 θˆ1 + 23 θˆ2 must be unbiased for θ. (b) Estimators obtained by the maximum likelihood method are unbiased. ¯ ·5 (c) If X1 , X2 , . . . , Xn are i.i.d. normal random variables N (µ, σ 2 ), then X P ¯ 2 /20, where X ¯ is the sample mean. is independent of ni=1 (Xi − X) 2. Let us have two independent random samples: X1 , . . . , Xn is a sample from N (µx , σx2 ), and Y1 , . . . , Ym is a sample from N (µy , σy2 ). (a) Write down a joint pdf for {X1 , . . . , Xn , Y1 , . . . , Ym }. (b) Find a 4-dimensional sufficient statistic. (c) Find the MLE of σx2 and σy2 . (d) Assume for this question only that σx2 = σy2 = σ 2 . Find the MLE for σ 2 . (e) Find a LR test statistic for testing H0 : σx2 = σy2 . (f) Suggest an exact test for testing H0 : σx2 = σy2 . Hint: if U ∼ x2p and V ∼ x2q are independent, then U/p V /q ∼ Fp,q (Fisher distribution) (g) Assume that n = 100, m = 50, s2x = 5, s2y = 6. Test the hypothesis H0 : σx2 = σy2 using the test received in (f) at 95% level. Tables of quantailes of Fisher distribution are provided. 6 σy2 by using an asymptotic LR test and (h) Test H0 : σx2 = σy2 vs. H1 : σx2 = the data as in (g). (i) Suggest a confidence set for β = 1 σx2 . σy2 MIT OpenCourseWare http://ocw.mit.edu 14.381 Statistical Method in Economics Fall 2013 For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.