Statistics is Easy New York University统计数据很容易纽约大学.pptVIP

下载本文档

9
0
约6.39千字
约 22页
2017-03-08 发布于上海
举报
版权申诉

Statistics is Easy New York University统计数据很容易纽约大学.ppt

1、本文档共22页，可阅读全部内容。
2、原创力文档（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。
3、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。
5、该文档为VIP文档，如果想要下载，成为VIP会员后，下载免费。
6、成为VIP后，下载本文档将扣除1次下载权益。下载后，不支持退款、换文档。如有疑问请联系我们。
7、成为VIP后，您将拥有八大权益，权益包括：VIP文档下载权益、阅读免打扰、文档格式转换、高级专利检索、专属身份标志、高级客服、多端互通、版权登记。
8、VIP文档为合作方或网友上传，每下载1次，网站将根据用户上传文档的质量评分、类型等，对文档贡献者给予高额补贴、流量扶持。如果你也想贡献VIP文档。上传文档

Statistics is Easy New York University统计数据很容易纽约大学

Statistics is Easy Dennis Shasha From a book co-written with Manda Wilson Is the Coin Fair? You toss a coin 17 times and it comes up heads 15 out of 17 times. How likely is it that coin is fair? Could look up Gaussian approximations to Bernoulli processes. Maybe you’ve forgotten… Or… Is the Coin Fair? You could do the following 10,000 times: toss a fair coin 17 times and count how many times you end up with 15 or more heads. Is this Practical? Takes under a second with your computer. Is this cheating? No, in the spirit of our times: solve differential equations with Euler’s method. Is it better? Yes, because more robust, easier to reason about, handles skewed distributions (e.g. average salary of $50,000 with variance of $15,000 gives non-zero probability of negative salary). What is the Result? Something like 9 out of 10,000 times, get 15 heads in 17 tosses. This gives a “p-value” of 9/10000. P-value is the probability that the outcome observed would have happened by chance if the coin truly were fair. Smaller p-value means less likely that the “null hypothesis” (coin is fair) is true. Is the Drug Effective? Random experiments are a good way to establish causality in the face of uncontrolled variation (e.g. lifestyle, wealth etc.) Suppose that we do an experiment in which we compare the positive effect of a drug vs. the effect of a placebo. Placebo: 54, 51, 58, 44, 55, 52, 42, 47, 58, 46 Drug: 54, 73, 53, 70, 73, 68, 52, 65, 65 Is the Drug Effective? Average for drug is 63.7 and for placebo 50.7. Is that the end of the story? Maybe not: perhaps the drug population “just happened” to be better. Standard statistical approach: assume something about distributions and see how big the overlap among distributions is. Issues with the standard approach? Distribution assumption may not hold. Must be careful about different sizes of one population vs. the other. Easy to get the technicalities wrong (for non-statisticians). Resampling/Shuffle Approach Create a table that a