Nonparametric Confidence Intervals For Regression Discontinuity Designs

In the process of the rapid development of our country and industry transformation,policies or rules need to be adjusted with the development of The Times,then will have an important influence on the development of related fields and industrial Whether it's positive or negative,it takes time to get a sense of what's going on.For the research perspective,many of these effects need to be analyzed and studied through relevant data.For the above ideas,a method of econometrics,namely Regression Discontinuity Designs(RDD),is highly compatible with research purposes not only in ideas,but also in practice.The main idea is that the observed amount is often divided into two groups according to whether the threshold is exceeded,and the group before the threshold is called the control group,and the group after the critical value is called the treatment group.The use of the threshold before the critical value is used as a control sample to reflect the impact of the policy on the key variable.With the passage of time,RD gradually received the attention of the people,and gradually formed a relatively fixed research idea.This paper is based on the RD theory.In the research process,the estimation of nonparametric local polynomial is widely applied to the interval estimation of the local average processing effect(LATE).However,in practical application,the obstruction caused by window wide selection is unavoidable.The existing bandwidth selection method is usually large,which will lead to the deviation of the confidence interval based on the data driven,so that the result cannot be convincing.Therefore,it is very meaningful to obtain the confidence interval estimation method with better practical effect.Correcting the deviation and using the new standard error estimation method construct a better confidence interval.The difference between this method and traditional method is that the new standard error estimator is applied.More intuitively,a quadratic nonlinear local regression estimator is used to rectify deviation.The results of simulation and empirical study prove that it works very well.