At developers meeting on 7/29 we discussed the following:
- The bug in OpenMx 1.1 Beta release where an ordinal non-positive definite covariance matrix goes into an infinite loop has been fixed.
- The 'connect' slot which is replacing the deprecated 'all' slot in MxPath is ready for testing. It will be checked in over the weekend for testing. For more information on this please see the forum post here.
- Changes and tests have been checked in for the PPML transformations. The tests are especially useful as they serve as a reference point and ground truth going forward.
- Speeding up the hessian calculation was discussed. There may be a different and faster implementation of the algorithm in the mle package. Exploring using the minor iterations of the npsol to reduce the number of computations and improve the efficiency was also discussed.
- Parallelizing objective functions was discussed. The team members decided the best parallelization strategy to explore first would be one that divided the data within a model. This choice was made because the strategy seems more straight forward to implement than the alternatives and it seems to be a slightly more common case parallelizing individual objective functions.
- A sequential strategy for improving the efficiency of the objective functions was also discussed: using derivative estimations. OpenMx developers should expect derivative-related literature via email in the upcoming week.
- The coding style of the models within the demos and the examples used in the OpenMx documentation was discussed. Background on this topic can be found in the 7/22 meeting notes available here. The recursive model style and the piecewise or object-add model style were both discussed in detail. Several team members identified the iterative and explicit nature of the recursive model style which enable it to be easily used in instructive settings. However, its lack of conciseness was also identified. Ultimately, the group decided that while it is important to show users demos and examples in different coding styles, no one style will be particularly pushed throughout OpenMx at this moment. An exercise where the different coding styles of each member of the team is shown on the wiki was discussed.
- Concerns about the lack of information/activity on the wiki were also voiced.