Bayesian Analysis Some Common Problems in Inference推荐.pdfVIP

8 Some Commo n Problem s in Inferenc e We have already discussed some basic inference problems in th e previous chapters. These include th e problems involving th e normal mean and th e binomial proportion . Some other usually encountered problems are discussed in what follows. 8.1 Comparin g Two Norma l Mean s Investigating th e difference between two mean values or two proportion s is a frequently encountered problem . Example s include agricultural experiment s where two different varieties of seeds or fertilizers are employed, or clinical trials involving two different treatments . Comparison of two binomial propor- tions was considered in Example 4.6 and Problem 8 in Chapter 4 . Comparison of two normal means is discussed below . Suppose th e model for th e available dat a is as follows. Yu^... ,Yin^ is a random sample of size ni from a normal population , N{9i^a\)^ whereas ^21, • • • 5 ^2n2 is an independent random sample of size n2 from another normal population , N{62,(72) ^1 ^ th e four parameter s ^1 , O2, cr^, and a l are unknown , but th e quantity of inferential interest is 77 = ^ 1 — ^2 • It is convenient t o consider th e case, cr^ = a l = cr^ separately . In thi s case, (Fi, 12,52) isjointly sufficient for ((9i, 62,cr2) where s^ = (E^liiYu - Yi)^ + E ; i i ( ^ 2 i - Y2)^)/{ni+n2 - 2) . Further , given (^1,^2,^) , 2 2 Fi ^ N{Qr. —).% - N{e2. — ) , and ( m + 712 - 2)s^ ^ cj^xl,^n,-2.