Improving linear quantile regression for replicated data
share
When there are few distinct values of the covariates but many replicates, we show that a weighted least squares fit to the sample quantiles of the replicates is asymptotically more efficient than the usual method of linear quantile regression.
READ FULL TEXT