|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑 8 [1 g: j1 W1 X b1 u9 k
煮酒正熟 发表于 2013-12-20 12:05 ) a! \# d2 J6 [4 q3 y9 B2 P9 ?2 l1 J
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 ...
! ]( W6 `5 h! d4 H; w: y; l
, D6 t$ k' P4 V y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)。
5 D5 J# P- ?4 o" g
7 x7 \% @+ |6 s6 {- ?8 @% H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)
" N" t. `& A4 l8 [
, K) M: B. `( e! D+ RR example:
. H' B9 E$ Q% W7 J2 n: q' P& s
# ~8 b% i# G; y- `3 I> M<-as.table(rbind(c(1668,5173),c(287,930)))
0 s# c- _' I* H: y+ L& r' T: U! D0 r> chisq.test(M)+ f# R9 c0 G" Q, r
4 Y9 d/ T5 X) j- n Pearson's Chi-squared test with Yates' continuity correction6 X2 g( `- b* [" d7 S
2 m! ?0 z0 E7 c
data: M# U2 Q/ c" u, J
X-squared = 0.3175, df = 1, p-value = 0.57315 q% j" P: I7 E% ?7 b
& q: S" X# n- ]# S% x0 ?
Python example:
% [, K" t1 ~8 x3 |, g! K
# A7 y$ g: J2 G! K! Y* ~. G7 s; |; u$ d>>> from scipy import stats0 _9 U: A9 ?7 x" Z) D/ q
>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])
0 [4 L4 a4 n5 z! |4 }(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],
& F8 n1 p. q# E% X3 W [ 295.26371308, 921.73628692]])) |
|