9. Apply the normal approximation (5.1)–(5.2) to Aspirin trial data (deaths di among myocardial infarction patients ni ) from Morris and Normand (1992, p. 334):

Compare the standard normal–normal model yi ∼ N(νi, σ2Vi ) i = 1, . . . , n νi ∼ N(μ, τ 2), with a robust alternative, namely yi ∼ N(νi, σ2Vi ) i = 1, . . . , n νi ∼ t(μ, τ 2, ν), with ν = 4, and using the scale mixture approach of Section 5.7. Are any outlier trials apparent?

Study Aspirin Placebo di ni di ni UK-1 49 615 67 624 CDPA 44 758 64 771 GAMS 27 317 32 309 UK-2 102 832 126 850 PARIS 85 810 52 406 AMIS 246 2267 219 2257