|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑 : b& u3 C V9 ^! w0 o6 q v& E
煮酒正熟 发表于 2013-12-20 12:05 ![]()
2 Z* L! |* C/ O. P8 F4 b) }5 A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 ... 6 q2 ^! j( I2 @ W, a8 w
4 J: r ~, l0 R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)。
$ g* H/ p1 i% ^, `5 z& x( z; E; s$ J/ M% g 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)! M" \7 p# a; X3 e" Z6 y
: O. R1 L$ Z+ E5 U( |. C, s
R example:9 L+ F# g9 G, }2 @( `
, C3 a# y8 @# I+ k: ^+ j1 X> M<-as.table(rbind(c(1668,5173),c(287,930)))
1 J" q' Q" O5 l; R$ K> chisq.test(M)
, {9 f$ R: [. K: d' R
0 e% n. b0 R; D6 c Pearson's Chi-squared test with Yates' continuity correction* m; m) X. I8 G* s" O: t
9 M& D9 k) G; c: H# R; w
data: M
6 b- b6 B1 \& T% O9 l4 nX-squared = 0.3175, df = 1, p-value = 0.5731
# [6 V# V3 o8 B( s7 ~
8 t* S7 x' \! h' Z; kPython example:
5 V4 k* z$ S5 b. {9 v" K l& s8 t6 m
>>> from scipy import stats3 q, i8 ]2 t3 C7 \ m4 Q u; A) y1 |( w
>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])3 Q8 L) A; |9 L/ l; r0 E: [
(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],
4 v7 e) |( U8 O! R* E+ j [ 295.26371308, 921.73628692]])) |
|