s*****w
发帖数: 90 | 1 我想要比较在哪个网站, A or B, 做广告更能提高产品的认知度. 请看下面数据.
在20个访问了A网站并且看到广告的人中(test group), 8个回答他们听说过我们的产品
, 在35个访问了A网站但是没有看到广告的人中(control group), 12 个回答听说过产
品. %lift b/w test and control is 17%. B网站的%lift是37.5%.
我现在想要比较这两个lift有没有显著差异. 请问大侠们有没有test可以用啊? 万分感
谢!!!
site A, test group 8 out of 20 voted yes (rate=40%); control group, 12 out
of 35 voted yes (rate=34%), %lift = 17% (40%/34%-1)
site B, test group 10 out of 30 voted yes (rate=33%); control group, 6 out
of 25 voted yes (rate=24%), %lift = 37.5% (33%/24%-1) | w*******9
发帖数: 1433 | 2 Haven't got an answer yet? You may try likelihood ratio test (where the
reduced model assumes lift1=lift2) or wald test (you can get the confidence
interval of lift1 and lift2 and thus the ci of lift1-lift2) | r*****t
发帖数: 91 | 3 可以用这个
if H1 (alternative hypothesis) is: p2>p1
test statistic z=(p2-p1)/sigma
sigma=sqrtroot(r(1-r)(1/n1 + 1/n2))
where r=(x1+x2)/(n1+n2)
x1 is the number of responses among n1 samples
x2 is the number of responses among n2 samples
p1=x1/n1
p2 = x2/n2 |
|
