Two proportion 检验使用之困惑
其实大家都知道Two proportion 检验是用来评定两组数据之总体比例是否相均等!本人使用一段时间对此方法略知皮毛,但还是对此有诸多不解之处!望各路豪杰勇于亮剑!具体参考数据见下:data-1.JPG通常我们会检验此两个比例是否相等,设定如下假设:!(1)设真值假设为proportion A=proportion B, 备选假设为proportion A not equal proportion B, 经过运算分析数据及p值如下: 样本 X N Sample p1 270 36820 0.0073332 5 2780 0.001799Difference = p (1) - p (2)Estimate for difference: 0.0055344195% CI for difference: (0.00373434, 0.00733448)Test for difference = 0 (vs not = 0): Z = 6.03 P-Value = 0.000 根据p-value=0.000<0.05可得出拒绝真值假设而接备选假设.(2)但如果用同样的真值假设 分别用另两个不同之备选假设, 2.1) 备选假设为proportion A less than proportion B, 2.2) 备选假设为proportion A greater than proportion B, 试着去分析运算数据及p值如下: 样本 X N Sample p1 270 36820 0.0073332 5 2780 0.001799Difference = p (1) - p (2)Estimate for difference: 0.0055344195% upper bound for difference: 0.00704508Test for difference = 0 (vs < 0): Z = 6.03 P-Value = 1.000样本 X N Sample p1 270 36820 0.0073332 5 2780 0.001799Difference = p (1) - p (2)Estimate for difference: 0.0055344195% lower bound for difference: 0.00402374Test for difference = 0 (vs > 0): Z = 6.03 P-Value = 0.000 由以上可知依不同的备选假设得出不同的结论,其中奥妙在哪?(原假设与备选假设要对立?)[ 本帖最后由 allen413 于 2008-6-3 20:57 编辑 ]