20/06/16 13:24:20.56 RYoHf/j0.net
>>692
多重比較になるけど、補正なしでも都市間の有意差はでないね。
むしろ、有名メーカーの検査キットの不一致の方が意外だった。
> positives=c(TYO=2,OSA=5,MYA=1)
> subjects =c(TYO=1971,OSA=2970,MYA=3009)
> pairwise.prop.test(positives,subjects,p.adjust.method = 'none')
Pairwise comparisons using Pairwise comparison of proportions
data: positives out of subjects
TYO OSA
OSA 0.82 -
MYA 0.71 0.21
P value adjustment method: none
Warning messages:
1: In prop.test(x[c(i, j)], n[c(i, j)], ...) :
Chi-squared approximation may be incorrect
2: In prop.test(x[c(i, j)], n[c(i, j)], ...) :
Chi-squared approximation may be incorrect
3: In prop.test(x[c(i, j)], n[c(i, j)], ...) :
Chi-squared approximation may be incorrect
> library(fmsb)
> pairwise.fisher.test(positives,subjects,p.adjust.method = 'none')
Pairwise comparisons using Pairwise comparison of proportions (Fisher)
data: positives out of subjects
TYO OSA
OSA 0.71 -
MYA 0.57 0.12
P value adjustment method: none