|
|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑
- w) z# y' I5 k& I( k' j, X煮酒正熟 发表于 2013-12-20 12:05 ![]()
2 v* ?( O- {- o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 ... & B( t' R! m/ v* x* G; c+ N
% ]0 @8 o: g% y% {4 U0 z ^
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)。 9 Z& c1 _: U4 U, i$ r8 ?& k3 {
4 A4 o5 j& |$ V, o
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)
+ c+ |9 M! T8 Q" }6 o; F! _# H& L0 |' G3 f
R example:
- T# Q# y3 }9 F1 J, u* ^0 O; c7 r4 s% X& e( @- r! W
> M<-as.table(rbind(c(1668,5173),c(287,930)))
1 _8 G- Y& E3 F% D, W> chisq.test(M)% E6 y; P1 \+ H# ^
; S6 q( n$ z' K+ [* S: ]. P0 B% _+ B
Pearson's Chi-squared test with Yates' continuity correction
" H7 b1 i+ j1 Q' E" o0 }4 f5 t: s6 Y" f, B4 q, }$ ]: X( l1 @
data: M
- u/ K) K0 J' R* H1 K3 aX-squared = 0.3175, df = 1, p-value = 0.5731
% d. _* L& Y( d+ j8 F3 D n0 N& y
Python example:& @$ P$ k. a& D4 p4 J1 K- {
7 M9 H5 U7 p8 M7 q% N0 `
>>> from scipy import stats. {/ R- z% S, m
>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])4 X$ Y/ z/ B: j% k
(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],
( G' ~) p: @0 a& [4 E* F [ 295.26371308, 921.73628692]])) |
|