Sparse Frisch-Newton Algorithms for Quantile Regression

Several recent developments in quantile regression are crucially dependent on sparse linear algebra to accelerate computations. In two papers Pin Ng and I describe modifications of the Frisch-Newton interior point methods proposed in an earlier Statistical Science paper with Steve Portnoy, that exploit sparse structure in the design matrix of the QR problem.
Inequality Constrained Quantile Regression,
A Sparse Frisch-Newton Algorithm for Quantile Regression,
Experimental R code for the Candes and Tao Dantzig selector is available here, A pure R version of the interior point LP solver is here,. This should be equivalent to the (much faster) fortran implementation of the same algorithm available in my quantreg R package. Note that if one specifies sparse = TRUE in the Dantzig function, then it will call the fortran version from the quantreg package.