범주형자료분석 개론

범주형자료분석 개론

범주형자료분석 개론.pdf

본문 (whitemales). Foralcohol,let S 1 =1,S 2 =0,andthenlinearpredictortakesgreatestvaluewhen R=1 andG=1(whitefemales). b. Race does not interactwith gender or substance type, so th e estimated odds for white subjects are exp(0.38)=1 .46timestheestimatedoddsforblacksubjects. c. Foralcohol,estimatedoddsratio= exp( .20+0 .37)= 1.19;for cigarettes,exp( .20+ .22)=1 .02; formarijuana,exp( .20)= .82. d. Estimatedoddsratio=exp(1 .93+ .37)=9 .97. e. Estimatedoddsratio=exp(1 .93)=6 .89. Theeectofalcoholuseonmarijuanause,controllingfo r cigaretteuse,seemstobeabitstrongerforfemalesthanfor males. 4. Similar substantive results occur. The GEE estimates, ba sed on an exchangeable working corre- lation,showalineartimeeectof0.318forthestandarddru gand0.318+0.708forthenewone. The dierenceinslopesof0.708hasaGEEempirical SE =0.136,sotheWaldstatisticequals(0.708/0.136) 2 =5.2( df=1, P < 0.0001)forcomparingtheslopes. 7. a. Subjects can select any number of sources, so given subj ect could have anywhere from 0 to 5 observationsinthe 40cellsofthis table. Multinomialdist ribution(andML ttingbasedonit)applies wheneachsubjectoccursinasinglecellofthetable. b. Estimated correlationis weak, so results not much diere nt from treating 5 responses by a subject asiffrom5independentsubjects. Thetablegivestheresult sfortheveequations,oneforeachsource, suchas 0.48750.2206sforthelogitforsourceD.ForsourceA,estimatedsizeeect is1.08andhighly signicant(Waldstatistic=6.46, df=1, P <. 0001). ForsourcesC,D,andEsizeeectestimatesare allroughly 0.2. 18. True. 하고 싶은 말 좀 더 업그레이드하여 자료를 보완하여, 과제물을 꼼꼼하게 정성을 들어 작성했습니다. 위 자료 요약정리 잘되어 있으니 잘 참고하시어 학업에 나날이 발전이 있기를 기원합니다 ^^ 구입자 분의 앞날에 항상 무궁한 발전과 행복과 행운이 깃들기를 홧팅 키워드 개론, 분석, 자료, 범주형, 범주형자료분석