|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑 3 M) A9 l) {$ O' q! t9 I" X1 ^. i+ _
煮酒正熟 发表于 2013-12-20 12:05 4 m" Z0 h0 M0 }6 ?
基本可以说是显著的。总的来说,在商界做统计学分析,95%信心水平是用得最多的,当95%上不显著时,都会去 ...
8 y/ i2 P: Q* X* U# [ E1 X
7 C6 h+ a* b7 C, M- ~; F& i这个其实是一种binomial response,应该用Contigency Table或者Logisitic Regression(In case there are cofactors)来做。只记比率丢弃了Number of trial的信息(6841和1217个客户)。
7 c6 k8 k% S& M7 C4 u) D5 Y, s9 H
' j) c2 q3 M# `6 y: `结果p=0.5731。 远远不显著。要在alpha level 0.05的水平上检验出76.42%和75.62%的区别,即使实验组和对照组各自样本大小相同,各自尚需44735个样本(At power level 80%)。see: Statistical Methods for Rates and Proportions by Joseph L. Fleiss (1981)
; R8 s2 _7 T4 C" s! g5 w2 c$ K! d+ J
R example:# C2 x( q4 B! a9 {
# h' p2 r: W: m2 Y; T: @
> M<-as.table(rbind(c(1668,5173),c(287,930)))
# [: R/ Q4 T) m# }+ F7 i> chisq.test(M)- V. A1 z5 H# s5 x! P6 F. A
$ F% E. ?1 ?4 ?( Z; [5 n Pearson's Chi-squared test with Yates' continuity correction; o9 W' B' w0 ~/ {
& u% Z4 W& i1 k, D3 Gdata: M
9 h/ f( P, A5 qX-squared = 0.3175, df = 1, p-value = 0.5731
, a! b/ [$ P0 F- N* W# F, \: @( A9 s
1 l3 Q; Q* c9 r9 X* YPython example:( ` S2 Z; K& k* v
" w/ P; Z2 P- k1 Q [+ ?
>>> from scipy import stats
! ~3 t5 ^7 E& A i>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])9 r: n7 b& v8 \
(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],& Y! j: G: d g# y2 K6 z+ R
[ 295.26371308, 921.73628692]])) |
|