南风金融网 - 中原最权威行情财经门户

热门关键词:  缪学刚  www.ymwears.cn  xxx    as
热门: 北大数字金融公开课内容回 对付勒索软件的利器:二级 还在为云存储掏冤枉钱?那 【Econ. for COVID-19文献6 【Workshop for MLinEcon第 直播预告 | “经世微语课堂

【Workshop for MLinEcon第3期】2020年4月21日

来源:南风金融网 作者:南风金融网 人气: 发布时间:2020-04-20 10:58:41

【题目1】:CombiningMachine Learning and Matching Techniques to Improve Causal Inference in ProgramEvaluation

【汇报人】:陶旭辉(上海财经大学公共经济与管理学院博士研究生)

【时间】:2020421日(周二)20:00-20:50

【地点】:Zoom在线

【论文摘要】Program evaluations often utilize various matching approachesto emulate the randomization process for group assignment in experimentalstudies. Typically, the matching strategy is implemented, and then covariatebalance is assessed before estimating treatment effects. This paper introducesa novel analytic framework utilizing a machine learning algorithm calledoptimal discriminant analysis (ODA) for assessing covariate balance andestimating treatment effects, once the matching strategy has been implemented.This framework holds several key advantages over the conventional approach:application to any variable metric and number of groups; insensitivity toskewed data or outliers; and use of accuracy measures applicable to allprognostic analyses. Moreover, ODA accepts analytic weights, thereby extendingthe methodology to any study design where weights are used for covariateadjustment or more precise (differential) outcome measurement. One-to-onematching on the propensity score was used as the matching strategy. Covariatebalance was assessed using standardized difference in means (conventionalapproach) and measures of classification accuracy (ODA). Treatment effects wereestimated using ordinary least squares regression and ODA. Results Usingempirical data, ODA produced results highly consistent with those obtained viathe conventional methodology for assessing covariate balance and estimatingtreatment effects. When ODA is combined with matching techniques within atreatment effects framework, the results are consistent with conventionalapproaches. However, given that it provides additional dimensions androbustness to the analysis versus what can currently be achieved usingconventional approaches, ODA offers an appealing alternative.

?

【文献信息】Linden, A., and Yarnold, R. P., “CombiningMachine Learning and Matching Techniques to Improve Causal Inference in ProgramEvaluation”, Journal of Evaluation in Clinical Practice, 2016, 22, 868-874.


【题目2】:Synthetic ControlMethods and Big Data

【汇报人】:王瑶佩(上海财经大学公共经济与管理学院博士研究生)

【时间】:2020421日(周二)20:50-21:40

【地点】:Zoom在线

【论文摘要】Manymacroeconomic policy questions may be assessed in a case study framework, wherethe time series of a treated unit is compared to a counterfactual constructed froma large pool of control units. I provide a general framework for this setting,tailored to predict the counterfactual by minimizing a tradeoff betweenunderfitting (bias) and overfitting (variance). The framework nests recentlyproposed structural and reduced form machine learning approaches as specialcases. Furthermore, difference-in-differences with matching and the originalsynthetic control are restrictive cases of the framework, in general notminimizing the bias-variance objective. Using simulation studies I find thatmachine learning methods outperform traditional methods when the number ofpotential controls is large or the treated unit is substantially different fromthe controls. Equipped with a toolbox of approaches, I revisit a study on the effectof economic liberalisation on economic growth. I find effects for severalcountries where no effect was found in the original study. Furthermore, Iinspect how a systematically important bank respond to increasing capitalrequirements by using a large pool of banks to estimate the counterfactual.Finally, I assess the effect of a changing product price on product sales usinga novel scanner dataset.

?

【文献信息】Kinn, D., “Synthetic Control Methods and BigData", Working Papers, 2018, 1803.00096, arXiv.org.

责任编辑:南风金融网

最火资讯

首页 | 财经资讯 | 金融理财 | 价格行情

豫ICP备12016580号  技术支持:南风金融网

电脑版 | 移动版