In high dimensions, sparseLTS() may be faster if the lasso on the subsets is computed via coordinate descent. For a grid of lambda values, warm starts can probably be used (i.e., use the sparse LTS solution from the previous value of lambda as initial coefficients in the lasso fits on the subsets).
In high dimensions,
sparseLTS()may be faster if the lasso on the subsets is computed via coordinate descent. For a grid oflambdavalues, warm starts can probably be used (i.e., use the sparse LTS solution from the previous value oflambdaas initial coefficients in the lasso fits on the subsets).