|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑 8 v! p6 d$ }9 p$ |
煮酒正熟 发表于 2013-12-20 12:05 " k$ V* z/ F3 ~8 ?( w5 O) Y# q) n
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 ... 5 C" W+ M( S8 _( ?
% p4 ~8 T( k. A9 [( B1 \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)。
8 G0 |3 ?. `' t/ p8 I) ^$ I! T+ c# f/ H
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)0 o+ V, ^+ B2 Q
- J# W, K6 F9 U- hR example:4 `$ \' Y" {6 ~4 t" ?2 N1 {$ Z z% q
8 ]- H; v( O+ }0 s# d% E& v> M<-as.table(rbind(c(1668,5173),c(287,930)))5 C [, k$ X% L1 ]) ^3 h1 @
> chisq.test(M)
) ^; a% g9 r! f5 Y& ~3 w. s ]0 ]# o; G! A6 _' |
Pearson's Chi-squared test with Yates' continuity correction$ g- z# I; Q. u7 K' @( f8 a
0 D2 `- m3 |+ { K: L4 O" B' Sdata: M2 b$ h7 F( k3 C& j6 I6 r3 q
X-squared = 0.3175, df = 1, p-value = 0.5731# e; s' _3 k; R, C ]% f2 w
! G% d! a" Z' }4 V. t- s! L
Python example:# a7 Y6 A0 Z/ \5 V( \5 Z
9 w" g- Q: N [. W, `>>> from scipy import stats
/ {! ]- @8 g8 H7 S; g B- J>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])( ?) ^( k. H. E
(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],
- W: Z7 U! x) B% h. M+ j4 x1 | K [ 295.26371308, 921.73628692]])) |
|