第九章配伍区组设计资料的统计分析例9-1某研究者用某药物治疗高血压患者10名,治疗前后舒张压的变化情况见表9-1。表9-110名患者用某药物治疗后的舒张压测定值(mmHg)患者编号(1)治疗前(2)治疗后(3)差值d(4)=(3)-(2)12345678910115110129109110116116116120104116901088992901101208896-120212218266-4328解:STATA数据为:x1x211151162110903129108410989511092611690711611081161209120881010496STATA命令为:ttestx1=x2结果为:Pairedttest------------------------------------------------------------------------------Variable|ObsMeanStd.Err.Std.Dev.[95%Conf.Interval]---------+--------------------------------------------------------------------x1|10114.52.1819976.900081109.564119.436x2|1099.93.89429812.3148591.09049108.7095---------+--------------------------------------------------------------------diff|1014.63.72737611.7876.16808923.03191------------------------------------------------------------------------------mean(diff)=mean(x1-x2)t=3.9170Ho:mean(diff)=0degreesoffreedom=9Ha:mean(diff)0Ha:mean(diff)!=0Ha:mean(diff)0Pr(Tt)=0.9982Pr(|T||t|)=0.0035Pr(Tt)=0.0018P=0.0035,治疗前后舒张压有差别,治疗后下降。例9-2某研究者对8名冻疮患者足部的两个冻疮部位(两个部位冻疮程度非常接近)用两种不同药物治疗,分别观测两个冻疮部位的痊愈时间,结果见表9-2。表9-2两种方法测定患者冻疮痊愈时间时间(天)结果受试者编号(1)药物1(2)药物2(3)差值d(4)=(3)-(2)181242109-1369344128576-161010078113810111解:STATA数据为:x1x2181221093694412576610107811810111.建立检验假设,确定检验水准00dH:,两种药物治疗的冻疮痊愈平均时间相同01dH:,两种药物治疗的冻疮痊愈平均时间不同05.0STATA命令为:ttestx1=x2结果为:Pairedttest------------------------------------------------------------------------------Variable|ObsMeanStd.Err.Std.Dev.[95%Conf.Interval]---------+--------------------------------------------------------------------x1|87.875.76619422.1671246.0632399.686761x2|810.707106828.32795811.67204---------+--------------------------------------------------------------------diff|8-2.1251.0763283.044316-4.670111.4201115------------------------------------------------------------------------------mean(diff)=mean(x1-x2)t=-1.9743Ho:mean(diff)=0degreesoffreedom=7Ha:mean(diff)0Ha:mean(diff)!=0Ha:mean(diff)0Pr(Tt)=0.0445Pr(|T||t|)=0.0889Pr(Tt)=0.9555t=1.9743,则P=0.0889,在05.0水平上不拒绝0H,差值的样本均数与已知总体均数的比较,差异无统计学意义,故尚不能认为该两种药物治疗的冻疮痊愈平均时间不同。例9-3为了解不同治疗方法对高胆固醇血症的疗效,根据专业要求,在采取相关清洗或洗脱措施,保证相邻两次疗效不受影响的前提下,某研究者用3种不同方法对9只受试动物进行实验,其血浆胆固醇测定值(mmol/L)见表9-3。表9-33种治疗方法的血浆胆固醇测定结果(mmol/L)动物编号甲方法乙方法丙方法12345678910.106.7813.227.787.476.116.028.087.566.695.4012.676.565.655.265.436.265.067.746.8310.957.206.855.885.797.876.45解:STATA数据为:bgx1110.1216.783113.22417.78517.47616.11716.02818.08917.56126.69225.43212.67426.56525.65625.26725.43826.26925.06137.74236.833310.95437.2536.85635.88735.79837.87936.45STATA命令为:anovaxgb结果为:Numberofobs=27R-squared=0.9378RootMSE=.672582AdjR-squared=0.8988Source|PartialSSdfMSFProbF-----------+----------------------------------------------------Model|109.0373281010.903732824.100.0000|g|11.125541425.562770712.300.0006b|97.9117867812.238973327.060.0000|Residual|7.2378600816.452366255-----------+----------------------------------------------------Total|116.275188264.47212262P=0.0006,3种不同方法得到的血浆胆固醇测定值(mmol/L)不全相同。例9-4将30只小白鼠按体重、性别、窝别、活泼性分成10个区组,每个区组的3只小白鼠随机分配到3个实验组,分别以不同蛋白质饲料进行喂养,60天后测量小白鼠的体重增加量(g),数据如表9-4。表9-4三种饲料喂养30只小白鼠的体重增加量(g)区组饲料ⅠⅡⅢ12345678930404141364833453733446252444149555377687681847875737410324872解:STATA数据为:bgx1130214031414141513661487133814591371013212332244326242525244624172498255925310248137723683376438153846378737583739374103721.建立检验假设,确定检验水准针对处理组H0:三种不同饲料喂养的小白鼠体重平均增加量相同H1:三种不同饲料喂养的小白鼠体重平均增加量不同或不全相同05.0针对区组H0:对于任何一种饲料喂养,10个区组的小白鼠平均体重增加量相同H1:对于任何一种饲料喂养,10个区组的小白鼠平均体重增加量不同或不全相同05.0STATA命令为:anovaxgb结果为:Numberofobs=30R-squared=0.9250RootMSE=6.00339AdjR-squared=0.8791Source|PartialSSdfMSFProbF-----------+----------------------------------------------------Model|7997.1333311727.01212120.170.0000g|7565.2666723782.63333104.950.0000b|431.866667947.98518521.330.2884Residual|648.7333331836.0407407-----------+----------------------------------------------------Total|8645.8666729298.133333则P处理0.01,P区组0.05,表9-6中P值为统计软件计算后直接给出的数值。可以说明,对于处理效应,按05.0水准,拒绝H0,可认为三种不同饲料喂养的小白鼠平均体重增加量不同或不全相同或至少有两个总体均数不同;对于区组,按05.0水准,不拒绝H0,还不能认为10个区组小白鼠的平均体重增加量不同或不全相同。例9-6某研究者采用1:1配对方法将16例肝炎患者分别分在两种不同治疗方法组,测定其血中GPT含量(iu/L),资料如表9-6第(2)、(3)栏,问:用不同方法治疗的患者GPT含量有无差别?表9-6不同治疗方法的肝炎患者血中GPT含量(iu/L)对子号(1)方法1(2)方法2(3)差值d(4)=(2)-(3)秩次(5)1112387462345678843017103233311247530622630697990-4577203-38451--3.557-23.5解:STATA数据为:x1x2111238284753303041762510326623330731698124791.建立检验假设,确定检验水准00dMH:,差值的总体中位数为001dMH:,差值的总体中位数不为005.0STATA命令为:signrankx1=x2结果为:Wilcoxonsigned-ranktestsign|obssumranksexpected-------------+---------------------------------positive|527.517.5negative|27.517.5zero|111-------------+---------------------------------all|83636unadjustedvariance51.00adjustmentforties-0.13adjustmentforzeros-0.25----------adjustedvariance50.63Ho:x1=x2z=1.405Prob|z|=0.15990.05P,所以没有足够证据可以拒绝H0例9-8某研究者欲了解不同受试者的血滤液在不同放置时间的血糖浓度,测定了8名正常人,将每位受试者的血滤液分成4份,然后随机地把它们放置0,45,90,135分钟,测定其血糖浓度,结果见表9-11。表9-11不同放置时间的血滤液所含血液浓度(mg%)受试者编号放置时间(分)045901351234567895(3.5)95(4)106(4)98(4)102(4)112(3.5)105(4)95(4)95(3.5)94(3)105(3)87(1)98(3)112(3.5)103(3)92(3)89(2)88(2)97(2)95(3)97(2)101(2)97(2)90(2)83(1)81(1)9