A/B Testing Udacity Notes

Dear Slim.

于 2021-12-26 16:46:35 发布

阅读量1.5k

点赞数

分类专栏：其他各种笔记文章标签：数据科学 A/B

本文链接：https://blog.csdn.net/weixin_43525427/article/details/122138851

版权

A/B测试用于比较现有产品和新版本的效果，适用于广泛而复杂的变化。不能用A/B测试来衡量新体验和错过的机会。测试周期包括细化假设、二项分布、置信区间、假设检验和样本量计算。它提供广泛的定量数据，而其他技术则提供更深入的定性信息。

摘要由CSDN通过智能技术生成

1. Definition

test set: existing products or feactures;
experiment set: the new version;
how customers response differently -> Which version of system better

2. Focus on all / specific changes?

A/B testing could generally used for broad and complicated changes you considering
<EXAMPLE: useful help you to climb up a curve mountain, but when you need to decide whether to climb up this mountain or that mountain, it may not be so useful>

3. What we cannot do with A/B test?

1.new experience;
2. what you missed?

reason: like/unlike, novolty effects, time costs
Question: 1. what’s your base comparison; 2. how much time we need til customer adapt to the new experience. (in the scope peroid, whether people recommen