Re: [機統] 請問一題統計
※ 引述《asaaaas (asaaaas)》之銘言:
: We consider a random sample X1,X2,...,Xn from a distribution with pdf
: f(x;θ) = (1/θ)exp(-x/θ) , 0<x<無限大 , zero elsewhere , where 0<θ.
: Possibly , in a life-testing situation, however ,
: we only observe the first r order statistics Y1<Y2<...<Yr.
: (a) Record the joint pdf of these order statistics and denote it by L(θ).
: ︿
: (b) Under these conditions, find the mle, θ, by maximizing L(θ).
: ︿
: (c) Find the mgf and pdf of θ.
: ︿
: (d) With a slight extension of the definition of sufficiency, is θ a
: sufficient statistic?
: (a)(b)小題已算出答案:
: r
: (a) [n!/(n-r)!][1/(θ^r)]exp[(-1/θ)(Σ yi + (n-r)yr)]
: i=1
: r
: (b) (Σ yi + (n-r)yr)/r
: i=1
: 想問(c)(d)這兩個小題
: (c)小題完全不會算
Y1 -> exp( n/θ) nY1 -> exp(1/θ)
Y2-Y1 -> exp( n-1/θ) --> (n-1)(Y2-Y1) -> exp(1/θ)
Y3-Y2 -> exp( n-2/θ) --> (n-2)(Y3-Y2) -> exp(1/θ)
Yr - Yr-1 ->exp( n-(r-1)/θ) + (n-(r-1))(Yr- Yr-1) -> exp(1/θ)
───────────────────
Σ yi + (n-r)yr -> Gamma (r,1/θ)
︿
θ -> Gamma (r,r/θ)
︿
mgf of θ= ( (r/θ) / (r/θ-t) )^r
︿ ︿ ︿ ︿
f(θ)=(θ^r-1) * (r/θ)^r * exp(-rθ/θ)/(r-1)!,θ>0
︿
: (d)小題我的想法是,用(a)和(c)算出來的結果,看f(y1,y2,..,yr;θ)/[pdf of θ]
: ︿
: 如果最後跟θ無關,則θ是充份統計,不知道這樣對不對?
: 我還有另一個想法,就是直接將(a)小題的[Σ(i=1~r)yi + (n-r)yr]這部份,
: ︿ ︿
: 換成rθ,則整個式子會變成只用θ和θ表示,沒有其他y出現,
: ︿
: 然後根據分解定理(Neyman),可知θ是充份統計。這樣對嗎?
: 但如果是用第二種想法,這樣(c)小題好像就沒意義了...
r
f(y1,y2,...yr;θ)=[n!/(n-r)!][1/(θ^r)]exp[(-1/θ)(Σ yi + (n-r)yr)]
i=1
--------------------------------------------------
r
令h(y)=1,T(y)= Σ yi + (n-r)yr),g(T(y);θ)= ↑
¯ ¯ i=1 ¯
f(y1,y2,...yr;θ)=h(y)*g(T(y);θ)
¯ ¯
由Neyman-Fisher分解定理得知 T(y) 為θ的充分統計量
¯
--
--
※ 發信站: 批踢踢實業坊(ptt.cc), 來自: 111.184.39.122
※ 文章網址: http://www.ptt.cc/bbs/Math/M.1407837525.A.6A7.html
討論串 (同標題文章)