應(yīng)yl7703永利官網(wǎng)趙學(xué)靖教授和李周平教授邀請(qǐng)，廈門(mén)大學(xué)林明教授將于5月31日至6月3日來(lái)我校交流并做學(xué)術(shù)報(bào)告，歡迎諸位老師、研究生參加！

報(bào)告題目：Estimating Counterfactual distribution functions via Optimal Distribution Balancing with Applications

時(shí)間：2024年6月3日(星期一)8:30

地點(diǎn)：理工樓631報(bào)告廳

報(bào)告摘要：To avoid estimating the inverse propensity weights, which is sensitive to model specification and easily causes unstable estimates, and often fails to adequately balance covariates in finite samples, this paper proposes a new estimator of counterfactual distribution functions. We find the weights of minimum dispersion that exactly balance the estimated conditional distributions among the treated, untreated, and combined groups via a well-defined convex optimization problem. The resulting estimator converges weakly to a mean-zero Gaussian process at the usual parametric rate $\sqrt{n}$, which does not need a high-order kernel in a nonparametric estimation. Also, we show that a properly designed Bootstrap method can be used to obtain confidence intervals for conducting statistical inferences, together with its theoretical justification. With the estimates of counterfactual distribution functions, we also provide the methods and theories to estimate the quantile treatment effects and test the stochastic dominance relationship between the potential outcome distributions. Monte Carlo simulations demonstrate that our estimator performs better than the inverse propensity weighting estimators in many scenarios. Finally, our empirical study revisits the effect of maternal smoking on infant birth weight.

報(bào)告人簡(jiǎn)介

林明，北京大學(xué)應(yīng)用數(shù)學(xué)博士，廈門(mén)大學(xué)經(jīng)濟(jì)學(xué)科教授，廈門(mén)大學(xué)經(jīng)濟(jì)學(xué)院副院長(zhǎng)，全國(guó)應(yīng)用統(tǒng)計(jì)專(zhuān)業(yè)學(xué)位研究生教育指導(dǎo)委員會(huì)委員。主要研究領(lǐng)域?yàn)橛?jì)量經(jīng)濟(jì)學(xué)、貝葉斯統(tǒng)計(jì)，現(xiàn)主持一項(xiàng)國(guó)家自然科學(xué)基金重點(diǎn)項(xiàng)目。

甘肅應(yīng)用數(shù)學(xué)中心

yl7703永利官網(wǎng)

萃英學(xué)院

蘭州大學(xué)大數(shù)據(jù)科學(xué)研究中心

2024年5月31日