|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑
( s8 t7 b0 W$ c+ g# `煮酒正熟 发表于 2013-12-20 12:05 ![]()
2 Q; s* [* o6 l" w! D2 w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 ...
9 ]/ S# ? A& z: y( n2 `* |+ a' i% k& ]
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 j& k; |' q" p: E" B
5 V' p) m, z2 V& t; ~! `+ R" U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)1 C) q8 j& x) a/ c& U/ q
, U0 U' q- @+ w+ F% B$ S# v1 @% d
R example:
0 m$ J/ g+ x5 H4 p9 J! W+ m7 F! T' e( w2 z4 `1 p9 s/ b8 v( `
> M<-as.table(rbind(c(1668,5173),c(287,930)))
) I }1 y5 F( @; H7 N, G3 Z1 q. _> chisq.test(M)# ]1 e% n) S, J: K8 g
2 V- N& S( }0 c0 q/ P
Pearson's Chi-squared test with Yates' continuity correction/ u1 v% m: c' l" B$ d
) k8 r- o, M, z+ l8 w, m3 E" \ mdata: M1 |: N! P* R& q
X-squared = 0.3175, df = 1, p-value = 0.5731' C" ~5 ]; `0 Z- o: u8 P
, F$ ` F) r/ {' k7 ?1 ^
Python example:/ v$ h$ s" n" b# [
7 k3 s; G3 J, ]1 ]4 V
>>> from scipy import stats
# p/ n! V# \: F5 |>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])
, Y# V4 l ^9 g6 X+ U0 U+ O(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],
r' c- P0 Y* B3 h/ _0 g [ 295.26371308, 921.73628692]])) |
|