|
|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑 2 t. N8 E4 b3 l
煮酒正熟 发表于 2013-12-20 12:05 $ |7 p' W5 C* f
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 ... ' m v1 v5 {, [9 y0 a2 v
; h2 Y5 Y! J$ K8 l& T$ _/ \
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)。
5 Q9 F* k! t- m9 h
0 g# d+ d: Q k& n3 z, N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)
/ ~' J1 M' \' w @
* Q# f" V- X$ T) T1 u$ hR example:
2 V6 ^; o2 {$ W' Y) P6 K
/ m" _6 g$ f+ H! k6 I> M<-as.table(rbind(c(1668,5173),c(287,930)))
3 g7 v1 |2 ~5 [+ z9 ~> chisq.test(M)0 s% R$ M5 p" O) L" n: t
, f, b6 | G# }2 ` Pearson's Chi-squared test with Yates' continuity correction. t8 b2 ^! U% ?4 e9 |
( O; n) n, r# \, M* J- r6 ^5 ^; S, U: }data: M
% S( K+ s. ^. w( y' n& ?$ ZX-squared = 0.3175, df = 1, p-value = 0.5731/ D, r' Y( \" t9 j- P* H M$ D
# V: U7 B. R2 x$ G9 @, Y
Python example:2 c( \% K7 A2 s( e8 v( }
; O; n% R# `* L ]
>>> from scipy import stats( P- l6 u$ Y' I, Z; p' D
>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])2 t( z) W- }0 O3 H6 p1 `
(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],
/ z$ B9 c6 C: x. L( @ [ 295.26371308, 921.73628692]])) |
|