|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑
) ]2 `/ b, I5 d煮酒正熟 发表于 2013-12-20 12:05 + t9 s* Y% S+ p& A# a" ?( I
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 ...
+ G: f# s; L$ ?6 B% l; C! T1 U# `; L" `* g, A; B' Q6 {8 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)。
3 c5 O# ?0 G$ w9 c( g
. z. j9 z0 P6 @, ]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), A2 b( V6 Q+ [' l: v% g
: @. p: F3 [7 |1 u- C% i2 jR example:
2 S! U0 ~; E0 m/ p2 e3 C) M, l4 f9 c& M+ X0 p
> M<-as.table(rbind(c(1668,5173),c(287,930)))0 X9 \5 C2 K$ `" k2 I% f
> chisq.test(M)
) K+ W; n0 W6 o; Y2 |9 R
" _' f7 V" S( i& A1 u/ L, ] Pearson's Chi-squared test with Yates' continuity correction
7 t) ^7 i, o! g/ _4 F; }) R2 B& I% h
data: M
: l) N k$ Y& |9 Z9 ?; fX-squared = 0.3175, df = 1, p-value = 0.57319 ^5 o' r/ j+ f+ ^) P- w7 D4 ?
0 G; B3 _- \0 f; z; D/ F2 o5 FPython example:
/ _0 S- B3 c, [' D G. [( s: A2 W5 u2 \1 B' I% a% g
>>> from scipy import stats- t+ D) Y/ O) V2 m0 X+ b, F6 z9 c
>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])
/ w4 u+ a" q/ x(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],! s7 U6 ]3 E# |$ W6 K- x! @
[ 295.26371308, 921.73628692]])) |
|