on mxOption()

No replies
carey's picture
Offline
Joined: 10/19/2009

this is a reply to mike spiegal's comments in the 'mxOption "Cold Start"' thread about the NPSOL options that should be implemented in OpenMx.

i suggest that eventually all the NPSOL options should be implemented, although i would not put this as a high priority in the developers hierarchy. the majority of users will have little knowledge of the inner workings of variable metric/quadratic programming optimization algorithms so mxOption() will be of little relevance. as one gets to more sophisticated problems (factor mixture models, simulations of natural selection using mixture fitness functions), however, tweaking the optimizer could really save time.

from my reading of the options given in the NPSOL manual, implementing them in OpenMx is less rocket science than tedious programming to verify that the arguments to mxOption() are legit.

hope other users can weigh in with their opinions.

best,
greg