數理統計LRT 的應用本頁導覽LRT 的應用 H0:θ∈ω0vs.H1:θ∈ω1H_0:\theta\in \omega_0\quad\text{vs.}\quad H_1:\theta\in \omega_1H0:θ∈ω0vs.H1:θ∈ω1 λ(x~)≜supθ∈ω0L(θ;x~)supθ∈ωL(θ;x~)∈[0,1]\lambda(\utilde{x})\triangleq \frac{\sup_{\theta\in\omega_0}L(\theta;\utilde{x})}{\sup_{\theta\in\omega}L(\theta;\utilde{x})}\in[0,1]λ(x)≜supθ∈ωL(θ;x)supθ∈ω0L(θ;x)∈[0,1] LRT reject H0H_0H0 ⟺ λ(x~)<k\iff\lambda(\utilde{x})<k⟺λ(x)<k Contingency table's Chi-square test Note: LRT reject H0H_0H0 ⟺ λ(x~)<k ⟺ −2lnλ(x~)>c\iff\lambda(\utilde{x})<k\iff-2\ln \lambda(\utilde{x})>c⟺λ(x)<k⟺−2lnλ(x)>c TheoremAs n→∞n\to\inftyn→∞, 在一些條件下−2lnλ(x~)→dχdf2∀θ∈ω0‾-2\ln\lambda(\utilde{x})\xrightarrow{d}\chi^2_{\text{df}}\quad\forall\theta\in \underbar{$\omega_0$}−2lnλ(x)dχdf2∀θ∈ω0 ω0‾\underbar{$\omega_0$}ω0 : 在 H0H_0H0 下發生 "===" 的 θ\thetaθ 的集合 df = dim(Ω)−dim(ω0‾)\dim(\Omega)-\dim(\underbar{$\omega_0$})dim(Ω)−dim(ω0