|
|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑 ; s z9 _$ j/ f6 E9 u
煮酒正熟 发表于 2013-12-20 12:05 % ~% R0 g' g: N+ I' E0 u
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 ...
4 l9 J/ z( Q( p3 `- R5 R% U R
) j0 p* K/ F4 n O4 x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)。 # K/ G* W9 C2 j/ }9 ?, T6 h
5 j. 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)
! I, H" R2 N; I! M+ @, P5 @& ?2 ]1 J# `2 Y# q7 O7 }
R example:. C+ c+ a- P: L7 F3 R: d
, J5 X& o, g" n
> M<-as.table(rbind(c(1668,5173),c(287,930)))( H U/ Q- L2 o' d8 e
> chisq.test(M)# y, y8 ?8 v+ k: F7 p8 l7 r
* i$ x5 P1 Z4 q' v# f4 X
Pearson's Chi-squared test with Yates' continuity correction
; p0 v8 Q* q6 }) m5 K7 ~' ?: c
+ ^- e% b2 W' [3 J6 h' G1 E( V6 B* ^data: M n! L$ l/ e$ J$ L, ?4 d
X-squared = 0.3175, df = 1, p-value = 0.5731
5 H q& W; D7 _4 ~0 T; }+ i7 }/ M! F) D8 |* F: ?0 i+ r2 |
Python example:
& W& C! Q' A$ g, j. D3 N
4 X: X, e: i9 T' b$ ^>>> from scipy import stats
8 p0 }$ Z4 n. Y>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])
; B6 }* n+ \1 T/ |0 T(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],7 k7 [. M& k P. J! m! P
[ 295.26371308, 921.73628692]])) |
|