Distributed inference for quantile regression processes (Q2414100): Difference between revisions

The big data sets provide a unique opportunity to discover subtle patterns in their distributions. The following method is proposed: (i) estimate the conditional quantile functions at different levels in a parallel computing environment; (ii) construct a conditional quantile regression process through projection based on these estimated quantile curves. \par The method divide-and-conquer is used: \par 1.) Divide the data with $N(x,y)$-pairs into $S$ subsamples of size $n$. 2.) Estimate the subsample based quantile regression coefficient. 3.) Each local machine sends the coefficient to the master that outputs the pooled estimator. \par As an important application, the statistical inference on conditional distribution functions is considered. Simulations confirm the considered statistical inference theory.