Numerous front-end additions were made that should be transparent to the user. Quite a bit of work was done on fixing standard errors. These are accurate fairly often now. However, it is not possible to cater for all cases, because the numerical accuracy of the fitted function varies considerably. There is a new option Err=x which allows changing the value of the function precision specifically for error calculations. By default it is set close to machine accuracy. Sometimes option err=1.e-10 (or larger) will correct problems with totally unrealistic estimates of standard errors.
Mike Neale, firstname.lastname@example.org Medical College of Virginia