ISDM seems to be in all sorts of trouble- look to see if I implemented this
on hsph; go back to L&K and try to improve
  - it doesn't converge on easy problems
  - if A is not of full rank it fails to work - it should!

All methods need to have the description section enhanced.

It would be nice to use Therneau's Surv class (but survival5 might be 
a lot of overhead).

Serious testing...

objects of class icsurv should indicate which fitting method was used.

it would be nice if we got MLEintvl to become MLEregion and have an abstraction
that worked in all dimensions....

there seem to be covariates available for the hiv data- we should include them
so that more complex analyses can be carried out

 somewhere I have a copy of Steve Self's data - that should get included.
