Issues with first beta release of OpenMx 2.0

1 reply [Last post]
RobK's picture
Offline
Joined: 04/19/2011

This thread is intended as a list of known issues with OpenMx 2.0 while it is in beta.

There are two known issues with the first beta, as of its public launch on 5/30/14:

  • This beta does not presently have an updated log of changes, so experienced users won't yet be able to get a quick summary of "what's different" by looking at the release notes (such as appear in Chapter 5 of the User Guide).
  • The new optimizer, CSOLNP, has a known memory leak: sometimes, using CSOLNP causes to R repeatedly allocate memory until it cannot do so anymore, which usually results in R crashing. CSOLNP is the default optimizer for the beta. However, the mxOption() function can be used to switch to using NPSOL if necessary. Nonetheless, we encourage those who are beta-testing version 2.0 to use CSOLNP as much as possible because it's a new feature.
  • CSOLNP can return inaccurate parameter estimates sometimes, too. Users should use NPSOL to verify results obtained with CSOLNP before trusting them!

Also, a word of advice: users should definitely get into the habit of ALWAYS using $ instead of @. For example, myModelRun@output$estimate would now be myModelRun$output$estimate. We CANNOT guarantee that every usage of the @ accessor that worked with versions 1.3/1.4 will continue to work in 2.0.

tbates's picture
Offline
Joined: 07/31/2009
toggling optimizers: mxOption(NULL, "Default optimizer", "NPSOL"

The currently selected optimiser can be checked with

mxOption(NULL, "Default optimizer")

To toggle the optimiser (currently "NPSOL" and "CSOLNP") you can say:

mxOption(NULL, "Default optimizer", "NPSOL")
# or
mxOption(NULL, "Default optimizer", "CPSOL")