Dependence-Robust Inference Using Resampled Statistics
30 Pages Posted: 15 Sep 2017 Last revised: 9 Jan 2019
Date Written: January 3, 2019
We propose new inference procedures robust to general forms of weak dependence. These procedures are based on a normal approximation showing that certain test statistics constructed using resampled data have standard normal limits. Implementation consists of comparing these statistics against an appropriate normal or chi-square quantile, depending on the statistic used. The statistics are simple to compute and do not depend on the correlation structure of the data. Furthermore, the normal approximation holds under the weak requirement that the target parameter can be consistently estimated at the parametric rate, which holds for regular estimators under many well-known forms of weak dependence. This justifies the claim of dependence-robustness. We consider applications to settings with unknown or complicated forms of dependence, with various forms network dependence as leading examples. We also develop a test for moment inequalities.
Keywords: resampling, dependent data, social networks, clustered standard errors
JEL Classification: C12, C31
Suggested Citation: Suggested Citation