|
本帖最后由 Menuett 于 2013-12-22 15:59 编辑 ; O; e7 Q% k( [3 ?
煮酒正熟 发表于 2013-12-20 12:05 4 [9 R& N$ l! O2 C/ ~1 X' k
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 ... ' D$ Z9 n! |3 \" A7 x8 U7 ^1 h. a
- M( A0 ^9 v- V% b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)。
: }! L/ o+ V( J, M# ~
# n- L* |7 m% M# t' c4 K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)" z+ P% t* C$ D3 n) ]
8 N% |# o3 H6 x" R0 `; uR example:
" L$ W+ z! T5 \9 E! N
6 w) C8 `' U c, b2 y5 R$ P> M<-as.table(rbind(c(1668,5173),c(287,930)))
7 }8 {; s9 h7 D- b> chisq.test(M)
3 x8 T; L& w9 i6 o7 R
# e" C4 n L0 Q& M) p" p Pearson's Chi-squared test with Yates' continuity correction
) `# \+ j+ D- v- M! ~' L' Q
+ T5 o+ @( p' h* X' f8 jdata: M+ t" q. f3 Q, l
X-squared = 0.3175, df = 1, p-value = 0.5731; f+ }( E8 _0 o5 q
/ `& A% N1 |/ q( b7 S m+ [# I
Python example:# B2 [% l q$ n! {' d
" @3 n& ?1 o" s$ V>>> from scipy import stats
" a+ V1 R1 w) r) b9 i! s>>> stats.chi2_contingency([[6841-5173,5173],[1217-930,930]])
) Z8 u+ }' q/ ~(0.31748297614660292, 0.57312422493552839, 1, array([[ 1659.73628692, 5181.26371308],. b. ?! w. C6 I! I* ]) x5 k
[ 295.26371308, 921.73628692]])) |
|