|
|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑
( T* Y" I& _0 D* v/ C! t( R! D( `煮酒正熟 发表于 2013-12-20 12:05 ![]()
& r; w' e* Q+ \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 ... $ |' X0 W8 D* t, K% m% ~7 C8 B/ P/ S
3 t( P2 r \+ 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)。 / s! v7 }3 Y; i8 U# H6 J/ u! ]
% o; J- h6 U" r" z y% V g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)
) q+ ^" Y# w* |; e" T% D
" z9 Q6 T7 T a2 F7 K4 E1 e) K- p; yR example:
8 `, y5 |9 P1 t3 a, u" k
- y% ?% j W5 Y1 D) s% i9 C' j> M<-as.table(rbind(c(1668,5173),c(287,930)))
: x0 u# u2 O' } l1 z: i6 r. g- s> chisq.test(M)
' @0 \. z `/ a+ E, u3 p& W5 V2 c$ G! a
Pearson's Chi-squared test with Yates' continuity correction( e) F' { }9 v) p% L
' H9 L8 I! c- w9 ldata: M
+ m. c2 j& d2 U2 d( AX-squared = 0.3175, df = 1, p-value = 0.5731
1 V& T6 m3 O5 R$ t; s, }8 }+ `5 a+ {: C- {/ v
Python example:
9 T: H% ~& O( \7 l @2 x) ?8 \+ U5 n: O) {/ l6 z
>>> from scipy import stats; x0 _8 I" P- H) o; R* A$ z
>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])6 c3 D2 j' D; [6 l$ I7 J
(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],
3 v" }( [; D9 j. G* e$ z [ 295.26371308, 921.73628692]])) |
|