|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑
* D ? n, @4 _3 l煮酒正熟 发表于 2013-12-20 12:05
& j) @) U9 |2 G& E/ R2 E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 ...
% P* d1 Y& ?3 n }) |! _2 ~/ h) _
% v6 u& Q( @( x0 E8 P# O1 D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)。
) T/ U+ _8 `7 u! i7 x9 Q
& W6 n% a& j b. |7 X3 ?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): k# c4 Q5 I: |3 R, [
2 M$ R9 x0 d$ v* {+ W4 O2 J7 f# b$ [
R example:
1 W# L) A5 k7 e" ?9 J
% I" I6 V/ D' p2 p4 k9 o' t> M<-as.table(rbind(c(1668,5173),c(287,930)))
* p4 U7 T2 J0 S) Z3 e) T( L' T> chisq.test(M)9 i# j& D2 m# d) {& h
, `4 K& a' n; j/ y; ? Pearson's Chi-squared test with Yates' continuity correction" ]9 K" m& a4 A; |7 o$ l4 y
$ l+ b ?5 e$ x0 kdata: M6 h3 h% g7 b, a# m% Y8 ~; ^+ H: P! D
X-squared = 0.3175, df = 1, p-value = 0.5731) r6 |! t: ^5 z5 m+ J9 @! O
5 P# Y/ J8 R4 ?- tPython example:
c# \; Q) u* }+ ^( m7 P" y" y, x) _
>>> from scipy import stats0 Q% `) Z p7 g; b5 R
>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])7 }. w" N* D$ d4 f3 W
(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],
/ C( B$ s3 t1 i. C [ 295.26371308, 921.73628692]])) |
|