Next: Comparison of Methods---I
Up: 5: Goodness of
Previous: Particular Robust Estimates
- Many forms for
have been proposed.
- The most successful was that due to Andrews with
- Note z is
where 
supplies
natural scale for problem.
- We do not need a very good estimate for
but it should be
robust!
- A simple choice is
which formula is applied iteratively as value of 
is
refined.
Geoffrey Fox, Northeast Parallel Architectures Center at Syracuse University, gcf@npac.syr.edu