Bootstrapping the process of model selection: AN econometric example
指出，研究者基于多次试估计选择模型后，最终结果的推断通常有误。作者将整个数据挖掘过程视为一个估计量，并展示自举法如何改进推断质量，并以死刑威慑效应的实证例子说明。
Abstract If a researcher has mined the data (i.e. selected an empirical model based on a series of trial estimates), inferences based on the final set of results are in general incorrect. This note treats the entire data mining process as an estimator and shows how a bootstrapping technique may improve the quality of inference. The method is applied to an empirical example on the deterrent effects of capital punishment.