多因子降维法和Logistic回归交互效应分析对比研究.pdfVIP

中淘药物与缅床2008年IO月第8卷第10期ChineseRemediesClinics．October2008，V01．8，No．10 ·777· 骆常好刘桂芬张蹙篷 i籀葵】 石c回归在不同条件下的适用情况。方法用MDR和Logistic回归实现不同条件下的交互作用分析。缡果MDR 可用以发现高阶交互作用．但不能发现主效应；Logistic回归能很好地分析主效应。甩在发现低阶交甄作用(二阶 交互作用)方面的能力比MDR更强大，但在分析高阶交互作用方丽有着很大的局限性。结论MDR朔Logistic 回归在分耩交互作用时有各自静适用条俘秘傀点。遇鸯交互作用分辑要求曩幸。二者的联合应用效果更镰。 【关键词l Logistic模蘩；圈l扫分析；多潮子降维法 ofinteractions reduction ￡C，O Analysis usi麟multifactordimensionalityVeli瞎U$I,ogistic黼伊蚓ond姗昭—；∞_ “UGui币mZHANG ofHealthStatistics，Shanxi 03000l，China Ai-lian．Department Medic越University,Taiyuan To the and withmuhifactor reduction [Abstract】Objectivecompareadvantagesdisadvantages dimensionality 8,8 eondition8for ofthe 撵Dand in ofinteractions。60to themostsuitable each Logisticregressionanalysis iustify BothMDRand were艄r|b册edforinteractionwithanumberof Methods approaches． Logisticregression analysis conditions．ResultsMDR usefulfordetectionof interactionsbutwasnotsensitiveto appeared hif；h-0rder identify andmore lo- maineffects．Ontheother in maineffects in hand．Logisticregressionw鹪adequaterecognizing capable showedalotoflimirationsin lOW—levelinteractionsftwO—factor MDR。but eating interactions)than analyzinghigh—or- derinteraetionConclusion霸b噍MDRand have and when Logisticregressionadvantagesdisadvantagescompared toeachother．Combined珏辩ofMDRand be when interaction