|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑
% _; b9 i: Q$ u% {0 [' Y/ V; X煮酒正熟 发表于 2013-12-20 12:05
( a7 }8 N. t; 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 ... / C0 K# L3 u0 A* D
( m5 ?. U# g. D6 S5 J! m
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)。
+ o' J" K- `2 Q3 |% }8 d4 F$ r+ i
: _6 w5 I- T( @7 d9 C H3 e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)) _8 ?7 l" w% n
) r9 Q8 L6 x5 P* {( e
R example:
r- a2 h9 a, `0 T6 ^$ X2 O% P& H% _
> M<-as.table(rbind(c(1668,5173),c(287,930))), T7 }! ^$ n4 o
> chisq.test(M)9 B0 Z: r1 F& Q- m' M
7 e/ O7 b9 Q- G2 `# f: w- s/ f3 e0 c Pearson's Chi-squared test with Yates' continuity correction5 m1 b% H: b& I: h% S
' I; ` }& h+ N+ K# T! C$ {data: M
9 }: M6 }5 K$ d3 ~# rX-squared = 0.3175, df = 1, p-value = 0.5731
7 ^/ {* a, G! p0 S5 A6 b, j- C' A+ T$ J' Y5 d0 a5 N2 K6 I
Python example:
5 b) Y9 }4 X! R9 Y3 f5 ?) _! q
" u& F, K: L. |' d>>> from scipy import stats, y; v9 J" t4 R- |$ I' w0 Q% E% i
>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])' i; P9 y, F; c( [) B7 S5 c: Q
(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],
- Z0 e4 P! r3 b6 T/ ? [ 295.26371308, 921.73628692]])) |
|