|
|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑 6 Y( z. p2 s( H+ k
煮酒正熟 发表于 2013-12-20 12:05 ![]()
' m2 ~; p k# A9 ?' [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 ...
1 r0 H7 N6 p# P0 _1 M
4 V4 X; q# B) H7 e6 X+ X% Q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)。 2 o, A( M4 O/ E3 v, T( F( \7 O
" A4 _# e2 {$ i. x$ x7 B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)5 q Q/ h# s; @! g; ] `
: ?5 G( `& N* ]+ z% z& c" }
R example:
- v. y$ m: E u ^
" R' X; r6 t% f- E. |> M<-as.table(rbind(c(1668,5173),c(287,930)))" n- w/ y, {; h
> chisq.test(M)4 s5 s& m9 m X" z8 K" F z
3 ?8 _. D& e1 Q/ a5 y5 `
Pearson's Chi-squared test with Yates' continuity correction
& C' d' p+ U1 B- J, R: C; w, ~; M
! J9 {$ ^( ~) rdata: M
1 D! L" _: B9 \7 u7 {X-squared = 0.3175, df = 1, p-value = 0.5731) g* W2 H$ m5 p& Z- X
1 ?$ D3 e T. H# d
Python example:; N0 Q9 b5 R! |( {
5 q% @8 K; a( M. n0 F/ [8 x3 Z# N
>>> from scipy import stats: B4 X0 U0 B; ~
>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])3 ~; z5 H; g& L$ n. I& U) v# k
(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],! a- q, u: o5 {! a8 S
[ 295.26371308, 921.73628692]])) |
|