應(yīng)用stata做meta分析之劑量反應(yīng)關(guān)系研究的寫書代碼第二章

OrsiniN,RuifengL,WolkA,KhudyakovP,SpiegelmanD.Meta- ysisforlinearandnon-lineardose-responserelationships:examples,anevaluationofapproximations,andsoftware.AmericanJournalofEpidemiology.2012;175(1):66-73.Summarizeddataabouttherelationbetweentotalalcoholintakeandlungcancerriskin4prospectivecohortstudiesparticipatinginthePoolingProjectofProspectiveStudiesofDietandCancer.第一步安裝GLST/數(shù)據(jù)sscinstalluse//Two-stagefixed-effectdose-responsemodelassumingglstlogrrdose,se(se)cov(peryearscases)pfirst(studytype)/*Relativeriskfor12grams/dayincrementalunit*/dose*12,eform個(gè)劑量時(shí)，的//Two-stagerandom-effectdose-responsemodelassumingglstlogrrdose,se(se)cov(peryearscases)pfirst(studytype)/*Relativeriskfor12grams/dayincrementalunit*/dose*12,eform第四步擬合非線性模型Non-linearityusingfixed-capturedrop_pctiledose,percentile(5356595)retlistmkspline = ,knots(`=r(r1)'`=r(r2)'`=r(r3)'`=r(r4)')cubicdisglstlogrrdoses*, cov(peryears pfirst(studyglstlogrrdoses1doses2doses3, cov(peryears pfirst(studytestparmdoses2//Figure1Bofthe logrrdose,se(se) cov(peryearscases) pfirst(studytype) predictnllrr_lin=_b[dose]*dosegenrr_lin=接下來計(jì)算的是以0DOSE//using0asglstlogrrdoses*,se(se)cov(peryearscases)pfirst(studypredictnllogrrwithref=_b[doses1]*doses1+_b[doses2]*doses2+_b[doses3]*doses3,ci(lohi)genrrwithref =exp(logrrwithref)genlbwithref =exp(lo)gen =接下來就是數(shù)據(jù)的列表了（非線性的情況）Tabulateresultusingxblc(finditxblc)levelsofdose,local(level)xblcdoses*,c(dose)at(`r(levels)')ref(0)twoway(linelbwithrefubwithrefrrwithref sortlp(longdashlongdashl)lc(blackblackblack))(linerr_lin sortlp(shortdash)lc(black)),scheme(s1mono)ylabel(.911.21.51.8,angle(horiz)format(%3.2fc))xlabel(0(5)45)legend(off)ytitle("RelativeRisk",margin(