|
|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑
" l' F, ]. b" D% z2 | w& e煮酒正熟 发表于 2013-12-20 12:05 ![]()
0 a! d- n3 V% l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 ... 0 J* v* a" t% H( C8 c9 M+ T2 P
* ]5 g* P$ p+ q& M' U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)。 - T: V5 b7 F7 M# K
/ I; X1 D8 V3 ^: i d, D, b
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)/ w& h' N8 |/ e% G
; ]9 u# `! W5 }- i* m1 BR example:
' R4 v( h+ _* \3 w- \3 t& j1 E3 G$ n8 y# L+ c
> M<-as.table(rbind(c(1668,5173),c(287,930)))
" z/ h- E# {1 ^$ X0 W4 f> chisq.test(M)
" ^( V, T) k1 l, f& N1 u# L
3 [; J7 P L A: N Pearson's Chi-squared test with Yates' continuity correction
1 W8 ^& {7 m( C8 [" S" Z6 F7 l* X6 x& B4 ^
data: M
" f6 H( e+ C7 z& q$ HX-squared = 0.3175, df = 1, p-value = 0.5731 C& a/ ]1 B: E+ C3 U: k- v& b
$ F0 Z( P$ {; c
Python example:
( q* j- p! t# O( u$ i
! \* x& y! ~% j; V( F; y>>> from scipy import stats2 P% z, ~' A5 ?/ n
>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])9 q+ H% J: X; M# A
(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],& \) [& G0 f2 V; G! H6 Y2 w
[ 295.26371308, 921.73628692]])) |
|