|
|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑
) \/ E6 k8 N& r6 @煮酒正熟 发表于 2013-12-20 12:05 0 i) {' n0 u% _8 n3 S0 K5 \1 B G
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 E, O7 I X4 q4 J
2 R& s% N1 K4 P5 q6 D& J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)。
& |/ Z& D. n( c/ W
2 v! z$ k$ N( T' Q9 `' B2 B& d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); o$ n, \; a+ E; U) i! m5 \* U
! e9 P2 s4 A& K3 b- s7 r
R example:
T, k) F* ]/ u
8 g J" B1 f+ {2 B# O5 T v> M<-as.table(rbind(c(1668,5173),c(287,930)))
' A0 c# @5 w3 @> chisq.test(M)
8 }$ l/ `; }7 S. S1 x, J) Y' k$ D8 [6 ^6 p$ }
Pearson's Chi-squared test with Yates' continuity correction
1 G- |2 i, d5 u5 e& o# n5 ?
2 i: ?3 c3 w5 d' u: qdata: M+ _ m1 S' k0 E# t+ X
X-squared = 0.3175, df = 1, p-value = 0.5731# j$ ^, ?! _6 _: M6 e
' f1 k; U9 }4 q
Python example:, M1 i. i3 n# E3 a
5 E) Z* n/ L9 W( c8 e>>> from scipy import stats
6 M% ?* B2 f/ z# k; ?+ H>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])
% `8 x6 j0 l7 J2 U$ e2 Y+ _6 E(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],# `- v. W& X( ]4 `/ }
[ 295.26371308, 921.73628692]])) |
|