|
|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑
. K. {2 r" V: }5 \+ g4 \% V煮酒正熟 发表于 2013-12-20 12:05 ![]()
" [- \9 {2 U& ~. 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 ...
6 s% e: H+ G6 W) r" g9 @( V1 u$ c0 E/ s8 T& a1 Y. l$ R% u2 C
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)。
! i0 y& R2 T( t8 z3 [+ Y
( G p+ s h4 q9 k& z" R; G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)
3 b- I. _2 t: |. O6 I
4 d, Z8 I" L3 e# ?R example:5 A3 z- v1 q; A7 S
2 \2 C1 S5 z/ }8 R6 {9 \7 ]
> M<-as.table(rbind(c(1668,5173),c(287,930)))5 O7 o; q" y9 w7 \. B
> chisq.test(M)+ g( T9 Y; l5 E
4 E! Z# T, D( O2 p" Q3 Z Pearson's Chi-squared test with Yates' continuity correction
8 \% B7 o/ S+ a$ A. O* f
2 m; l0 a4 j! b1 p6 P$ sdata: M
?0 y/ l6 O$ ?# I6 PX-squared = 0.3175, df = 1, p-value = 0.57317 n6 C+ M, S" ` D$ M! \& U. @' H
7 J* u! D9 L( ?" V. l" E R) }Python example:
" F% T2 L) `' ~3 n) W* _1 E6 m1 }9 x) K
>>> from scipy import stats
; z( J& E7 x$ G4 G>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])
- e( t* K- \" U+ |(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],& o0 F2 H( q, m6 _
[ 295.26371308, 921.73628692]])) |
|