At the Developers' Meeting on 4/20/12 we discussed the following:
- The group discussed the urgency for two features: (1) weighted least squares and (2) multi-level modeling. The interfaces changes to accomodate WLS and other fit functions have been decided and implementation should start soon. It appears the best time to start implementing multi-level is in August while a quorum of group members will be together.
- Steve has continued to investigate the RODBC package. This investigation has brought to light the need to document our recommendations for how we should structure data hierarchical/relational data in OpenMx. Once we have done that we can start making adjustments to the RODBC package to meet our needs.
- The group discussed the early timing results for Csolnp (the C translation of the solnp optimizer). For the Powell problem benchmark it is two orders of magnitude (~200x) faster than the R version of the solnp optimizer. This is to be expected due to the differences in C and R.
- Mike Hunter has developed a manual page for the implementation of LISREL models in OpenMx. Shortly, he will be publishing a forum post that invites users to experiment and test LISREL models, using the direction of the manual, and report any errors or hiccups they may have to the wiki.
- The group discussed future NSF funding opportunities for OpenMx.