|
|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑 3 b) u: t( |- {% I
煮酒正熟 发表于 2013-12-20 12:05 8 r8 Q" Z0 k, `8 S* z
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 ... , h+ L+ Q5 s$ X% K1 X( [$ f# \
j: }+ R" N5 i- p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)。
' E0 \7 |0 \5 U ]5 M7 P7 ^9 N7 a/ y% |4 r8 b# s" K0 G# ^: {) i5 i
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)
! c1 ^- G# [, N: ^0 R( d
3 H, T2 n& a' M/ VR example:
. `$ F. u$ u7 n( g- M4 K% a2 |- X8 I3 e" p- D+ u
> M<-as.table(rbind(c(1668,5173),c(287,930)))5 K- M5 s* ^. x) ? P
> chisq.test(M)# f2 L: @+ T% y/ P
4 j8 U& i, \. ~ i
Pearson's Chi-squared test with Yates' continuity correction
. b& h# M( L7 t1 Y' B$ @' K9 P8 c: Q- a, C" J
data: M
# U) F/ o W6 DX-squared = 0.3175, df = 1, p-value = 0.5731
3 {. ]7 s C& e8 |
/ n; N( F5 R2 V& t) n& A& l8 \Python example:* D4 H+ Z% F- X) u5 s* A1 l
/ U- }( C L! x3 ?>>> from scipy import stats
" L9 B1 ]$ ]9 k6 j>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])7 C, I m4 ?( P7 a
(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],0 ~& G. ~( m6 U/ v
[ 295.26371308, 921.73628692]])) |
|