A Corrected Plug-In Method for the Quantile Confidence Interval of a Transformed Regression

Abstract: In this paper we propose an analytically corrected plug-in method for constructing confidence intervals of the conditional quantiles of a response variable with data transformation. The method can be applied to (i) a general conditional regression quantile, (ii) a general monotonic transformation, and (iii) a transformation model with heteroscedastic errors. Our results extend those in Yang (2002a), in which the median of a response variable under the Box-Cox transformation with homoscedastic errors was considered. A Monte Carlo experiment is conducted to compare the performance of the corrected plug-in method, the plug-in method and the delta method. The corrected plug-in method provides superior results over the other two methods