狗家 QA

发信人: sbw8687 (grit), 信区: Statistics
标  题: 狗家 QA
发信站: BBS 未名空间站 (Mon Dec 22 10:47:59 2014, 美东)

电面几个题目
1. a coin, if we want to know whether it is biased, how many times we need
to toss it?
2. build model to predict whether a customer will click a link? (could use
logistic regression)
what should we do if one customer contributes to several observations, i.e.,
repeat measures?
3. build model to predict how many times a customer will visit a web page? (
could use poisson model), how if the model fit is poor, any potential
problem with poisson model? how if a lot of zeros are observed?
4. a very long array (can not hold by the memory), how to calculate its mean
? (my answer: partition the array)
how to calculate its variance? if we only want to visit each element of the
array once (to calculate the variance), what should we do?

每个题目大概都有几个subquestion，所以考虑问题还是要全面一些。。。
--
※ 来源:·WWW 未名空间站 网址：mitbbs.com 移动：在应用商店搜索未名空间·[FROM: 130.]

http://www.mitbbs.com/article_t/Statistics/31380747.html

FT中文十大 华尔街中文十大 水木十大 小百合十大 饮水思源十大 日月光华十大

京东极速达北京支持货到付款 ， 更多京东优惠券