《Logistic回歸分析(重點、難點)【行業(yè)一類】》由會員分享，可在線閱讀，更多相關(guān)《Logistic回歸分析(重點、難點)【行業(yè)一類】（65頁珍藏版）》請在裝配圖網(wǎng)上搜索。
1、Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦掌握掌握了解了解教學目標教學目標理解理解幾個重要的流行病學概念幾個重要的流行病學概念Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦幾個重要的流行病學概念幾個重要的流行病學概念v隊列研究（隊列研究（cohort studyc
2、ohort study）v病例對照研究（病例對照研究（case-control studycase-control study）v危險度（危險度（riskrisk）v相對危險度（相對危險度（relative riskrelative risk，RRRR）v比數(shù)比或優(yōu)勢比（比數(shù)比或優(yōu)勢比（odds ratioodds ratio，OROR）Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦隊列研究隊列研究Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦病例對照研究病例對照研究Depa
3、rtment ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦危險度危險度 Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦相對危險度相對危險度 Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦相對危險度相對危險度 )/()/(dccbaaRRDepartment ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦比數(shù)比比數(shù)比Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學
4、教研室 張彥琦張彥琦比數(shù)比比數(shù)比bcaddcddccbabbaaOR/)/()/(/)/()/(ORbcaddcbadccbaaRR/)/()/(bcadOR/Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦一個病例對照研究資料一個病例對照研究資料Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦 什么情況下采用什么情況下采用LogisticLogistic回歸回歸Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦什么情況下采
5、用什么情況下采用LogisticLogistic回歸回歸Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦用途：用途：研究某種疾病或現(xiàn)象發(fā)生和多個危險因素（或研究某種疾病或現(xiàn)象發(fā)生和多個危險因素（或保護因子）的數(shù)量關(guān)系。保護因子）的數(shù)量關(guān)系。用用 檢驗（或檢驗（或u u檢驗）的檢驗）的局限性局限性：1.1.只能研究只能研究1 1個危險因素；個危險因素；2.2.只能得出定性結(jié)論。只能得出定性結(jié)論。2什么情況下采用什么情況下采用LogisticLogistic回歸回歸Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室
6、衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦什么情況下采用什么情況下采用LogisticLogistic回歸回歸Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦實例實例Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦實例實例Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦實例實例Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦實例實例Department ofHealth Sta
7、tistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦)1ln()(PPPLogit1 0 Y發(fā)生應變量未發(fā)生12,mXXX自變量),|1(21mXXXYPP1P0)(PLogitDepartment ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Z)(exp11PLogitPDepartment ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦00.51-4-3-2-10123410.5PZ:,0,:0,0.5,1PZ Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室
8、張彥琦張彥琦),2,1(mjjjXmmXXXPPPLogit2211)1ln()(jXDepartment ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦1100/(1)/(1)jPPORPPDepartment ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦111000010010/(1)lnlnlogitlogit/(1)()()()jmmjttjtttjtjjPPORPPPPcXcXcc對對比比某某一一危危險險因因素素兩兩個個不不同同暴暴露露水水平平1cXj與與0cXj的的發(fā)發(fā)病病情情況況（假假定定其其它它因因素
9、素的的水水平平相相同同），其其優(yōu)優(yōu)勢勢比比的的自自然然對對數(shù)數(shù)為為:10 exp()jjORcc即Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦0,1 exp,0,1 0,1 jjjjjjOROROROR無作用危險因保子護因子則有101 ,1,0 jccX暴非若暴露露10 exp()jjORcc即Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦1100/(1)1,/(1)PPPORRRPP當則有由于jOR值與模型中的常數(shù)項0無關(guān)，0在危險因素分析中通常視其為無效參數(shù)。Depar
10、tment ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦 11(1)iinYYiiiLPP 1lnln(1)ln(1)niiiiiLYPYP mbbbb,210 Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦)(exp01ccbROjj若自變量jX只有暴露和非暴露兩個水平，則優(yōu)勢比jOR的1可信區(qū)間估計公式為)exp(2/jbjSubDepartment ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦2,1jjjjbbbbuSS2 或 Wald 檢驗 將各參
11、數(shù)的估計值jb與 0 比較，而用它的標準誤jbS作為參照，檢驗統(tǒng)計量為 2Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦vData Weight casesDepartment ofHealth Statistics衛(wèi)生統(tǒng)
12、計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦vAnalyze Regression Binary LogisticDepartment ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Dependent Variable EncodingDependent Variable Encoding01Original Value.001.00Internal ValueVariables in the EquationVariables in the Equation.546.15811.9891.0011.727-.149.1091.8511.174.862吸煙史C
13、onstantStep1aBS.E.WalddfSig.Exp(B)Variable(s)entered on step 1:吸煙史.a.Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦實例分析實例分析Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Department
14、ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦實例分析實例分析Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦條件條件LogisticLogistic回歸的原理回歸的原理Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦表表16-5 1:M 條件條件logistic回歸數(shù)據(jù)的格式回歸數(shù)據(jù)的格式*t=0 為病例，其他為對照 Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦條件條件Logis
15、ticLogistic回歸模型回歸模型 0112211 exp(.)iimmPXXX 1,2,in與非條件 logistic 回歸模型不同之處在常數(shù)項上，不同匹配組的i0可以各不相同，但內(nèi)在假定了每個危險因素的致病能力在不同匹配組中相同。Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦實例分析實例分析因素因素變量名變量名分級說明分級說明不良飲食習慣不良飲食習慣X X1 10 0、1 1、2 2、3 3、4 4表示程度（表示程度（0 0表示無，表示無，4 4表示最多）表示最多）喜吃鹵食和鹽喜吃鹵食和鹽漬食物漬食物X X2 20 0、1 1
16、、2 2、3 3、4 4表示程度（表示程度（0 0表示不吃，表示不吃，4 4表示非常表示非常喜歡吃）喜歡吃）精神狀況精神狀況X X3 30 0表示差，表示差，1 1表示好表示好Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦v數(shù)據(jù)準備成4列：X1、X2、X3，配對號及疾病發(fā)生情況Y（1：病例，0：對照）。實例分析實例分析Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦v 模型擬合結(jié)果Variables in the EquationVariables in the Equatio
17、n.991.30710.4081.0012.693.901.3317.4031.0072.462.505.534.8941.3441.657X1X2X3BSEWalddfSig.Exp(B)v 顯然，存在不顯著的變量，可以進行逐步回歸分析。實例分析實例分析Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦V Va ar ri ia ab bl le es s i in n t th he e E Eq qu ua at ti io on n.991.30710.4081.0012.693.901.3317.4031.0072.462.50
18、5.534.8941.3441.657.945.29410.3551.0012.572.882.3247.4041.0072.416X1X2X3Step1X1X2Step2BSEWalddfSig.Exp(B)實例分析實例分析第三節(jié)第三節(jié) LogisticLogistic回歸的應用回歸的應用 及注意事項及注意事項Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研
19、室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦)ln(exp110XPDepartment ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦目的：目的：作
20、出以多個自變量（危險因素）估計應變量（結(jié)果作出以多個自變量（危險因素）估計應變量（結(jié)果因素）的因素）的logistic回歸方程。屬于概率型非線性回歸?；貧w方程。屬于概率型非線性回歸。資料：資料：1.1.應變量為反映某現(xiàn)象發(fā)生與不發(fā)生的二值變量；應變量為反映某現(xiàn)象發(fā)生與不發(fā)生的二值變量；2.2.自變量宜全部或大部分為分類變量，可有少數(shù)數(shù)自變量宜全部或大部分為分類變量，可有少數(shù)數(shù)值變量。分類變量要數(shù)量化。值變量。分類變量要數(shù)量化。用途：用途：研究某種疾病或現(xiàn)象發(fā)生和多個危險因素（或保研究某種疾病或現(xiàn)象發(fā)生和多個危險因素（或保護因子）的數(shù)量關(guān)系。護因子）的數(shù)量關(guān)系。種類種類:1.1.成組（非條件）成組（非條件）logistic回歸方程。回歸方程。2.2.配對（條件）配對（條件）logistic回歸方程?；貧w方程。Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦)(exp112211mmXXXP模型：模型：LogisticLogistic回歸模型是一個概率型的回歸模型。回歸模型是一個概率型的回歸模型。應用：應用：Department ofHealth Statistics衛(wèi)生統(tǒng)計學教研室衛(wèi)生統(tǒng)計學教研室 張彥琦張彥琦
Logistic回歸分析(重點、難點)【行業(yè)一類】
