ST562 Part3 - pt2
ST562 Part3 - pt2
docset apend on B 9B on oncttany spe Let X1,X be tld gamma(a. 3). Let Z = jpiieatt8)_ODavid Soave 2002 Page 15 Complete Statistics twitively, a minimal sulficlent statistic eliminates all the extraneouis information x the sample. Thus, we would suspect that a minimal sufficient statistic is inde~ encdent of (unrelated to) an ancillary statistic. 9 quan. gulf. toe ye. UHV «+ For many important situations this is true! a gt « Note, however, there are situations where this is not the case: The Definition 8.6 Let Jt) bea family of pals or pf for atatsle 11) ee family of probability distributions is called complete if Eo|9(7)) = 0 1 istic PolglT = | for all , Equivalently, T(X) is called a complete tal obability distributions, + Notice that completeness is a property of a family of Pr not of a particular distribution. X, be Family Let Xi tial Theorem 3.5 Complete Statistics in the Expone! fof the form pf or pn Aid observations from an exponential family with pa oF P Sxl) = hz)e(@exp where = (0,,...,0y). Then 11x) = (= ) Susi) i yt ‘sa complete statistic i {(u4(b),...,ws(@)) 0} contains an open set in 1. * Sometimes we wall refer to these types of exponential families as nice Example 3.14 bet X00. X isa comple! be ld N(y2.02) where a? Is known, Show that the sample mean 7(.) = sufficient st Istle for jp, exp(tatygaatyy FRU = Lene8) CB. ote et ear Sle any 1 eee a 1h as casa ewes Tew Ear compete cutfcest for Page 18 o 361 (Part 3) _@David Soave 2022 _ Exomple 3.16 Let X), X,, be lid observations with pdf log)" yen c1,0> JS(2\0) = (oat Find a complete sulficient statistic for . loge + @>1 , 1098 > 0 > Open ser OR 2 Tew > T 51S complete, sufficient for O