You can specify the COVB and CORRB options in the MODEL statement to request covariance and correlation matrices for the estimated parameters. The QUANTREG procedure computes these matrices only for a single quantile at a time.
The QUANTREG procedure provides two methods for computing the covariance and correlation matrices of the estimated parameters: an asymptotic method and a bootstrap method. Bootstrap covariance and correlation matrices are computed when resampling confidence intervals are computed. Asymptotic covariance and correlation matrices are computed when asymptotic confidence intervals are computed. The rank method for confidence intervals does not provide a covariance-correlation estimate.
Asymptotic Covariance-Correlation
This method corresponds to the sparsity method for the confidence intervals. For the sparsity function in the computation of the asymptotic covariance and correlation, the QUANTREG procedure provides both i.i.d. and non-i.i.d. estimates. By default, the QUANTREG procedure computes non-i.i.d. estimates.
Bootstrap Covariance-Correlation
This method corresponds to the resampling method for the confidence intervals. The Markov chain marginal bootstrap (MCMB) method is used.
