Testing of rq, R version:

A. Running the examples in rq help:
   rq(stack.x,stack.loss,.5) 
              R gives warning about possible non-unique solution, which S
              doesn't.  But the computed solutions are the same, to full
              precision printed.  Warning might have to do with conversion
              to double precision?

   rq(stack.x,stack.loss,tau=.5,ci=T,method="score") 
              R gives warning about possible non-unique solution, not S. 
              as above.

   rq(stack.x,stack.loss,.25)
              Help comments: note that 8 of the 21 points lie exactly 
	      on this plane in 4-space.  For the R-solution, only 4
              points! $h component of solution is different! This is the
              index of the observations in the basis, so indicates that
              indeed different solutions have been found. But coefficients
              and confidence intervals are identical.  

   rq(stack.x,stack.loss,-1) 
             As above, some differences in $h component. 

   rq(y=rnorm(10),method="sparsity")
             OK.


