Data Scientist Intern面试两问

生统专业，可以套sas的数据处理吗？epi的知识用回答第三题合适吗
1.How do you approach a completely new set of unknown data? Where do you start?
2.How can you tell if your conclusions are valid? What have been the most typical reasons for invalid conclusions?
没有多少hands-on 经验的菜鸟求教地里大神

sas? i think these questions have nothing to do with software but "approach"
首先是看看数据的分布什么的吧? discrete or continuous etc
结论是否正确，看 performance on a hold out test set? (cross validation)

invalid conclusion: hmm, current data set does not well represent the underlying population,  or it could be that your model doesn't fit data well (many many scenarios), other explanatory factors that were not included in the available data, training set too small, wrong model, lack of fit,

what else?

小K 发表于 2013-3-9 07:44
sas? i think these questions have nothing to do with software but "approach"
首先是看看数据的分布什 ...

谢谢小k，一定要别人指点思路才有信心。。。

继续讨论吧，我觉得我第二题回答的也不是很好诶。。。

第一题，应该还需要知道这个数据是用来做什么的吧，main objective是什么，数据是怎么收集上来的，能不能用来回答想问的问题

小K 发表于 2013-3-9 07:59
第一题，应该还需要知道这个数据是用来做什么的吧，main objective是什么，数据是怎么收集上来的，能不能用 ...

没有实干过没概念呀，第一步要不要进行data clean，比如missing value 怎么code的，发现outliers要不要排除，但是这个思路是不是太细了根据需要来的

周二有个面试， 感谢分享~

求问楼主这个intern是通过什么途径找的？

都是宝贵经验，Mark下