optim options
Closed this issue · 5 comments
looks like we have the option lower used in many of the optim() statements which isn't compatible with all optim methods. We should try to optimize on a transformed scale when possible and get rid of lower (and upper) options.
yeah so other optim methods really don't work. Unless we fix this we may want to remove opt.method or say that it doesn't work in the documentation.
Or the option in but have it throw a message for any other methods saying "only XXX is supported currently". This seems like it would be simplest later on if we do figure out how to add them.
Nicholas G. Reich, Ph.D.
Research Assistant Professor
Division of Biostatistics and Epidemiology
School of Public Health and Health Sciences
University of Massachusetts, Amherst
413.545.4534
nick@schoolph.umass.edu
http://people.umass.edu/nick/
on twitter: @reichlab
On Tuesday, July 9, 2013 at 9:54 PM, Andrew Azman wrote:
yeah so other optim methods really don't work. Unless we fix this we may want to remove opt.method or say that it doesn't work in the documentation.
—
Reply to this email directly or view it on GitHub (#3 (comment)).
Other methods are working now but clearly they won't all work always. Need to now make sure we have a reasonable error that suggests changing the opt.method
From Justin: "I think we should just throw an error if we cannot fit, and suggest they try different starting parameters or an alternate optimization method? If the data cannot be fit there is not really anything we can do about it."