You are here

Version 2.7.12 of OpenMx officially released

We are pleased to announce the official release of OpenMx version 2.7.12. Click here for instructions on how to install the package from our repository. As usual, our repository has package binaries for Windows and MacOS, and source tarballs for Linux/GNU; our builds of the package include the NPSOL optimizer.

Users of Windows and Linux/GNU can also install OpenMx v2.7.12 from CRAN. However, CRAN has not been able to build functioning OpenMx binaries for MacOS since OpenMx v2.7.10. MacOS users can still obtain version 2.7.12 via CRAN, but the Mac onto which you are installing the package must be configured to build it from source. Again, precompiled MacOS binaries are available through our own repository.

New features, performance improvements, and bug-fixes in OpenMx v2.7.12 include the following:

  • Previously, in multilevel models, the Rampart sufficient-statistic speedup did not work correctly with path models involving zero-variance exogenous covariates, which led to numerically incorrect results.
  • We have adjusted the "threshold" on the gradient norm, at the optimizer's solution to an unconstrained problem, above which OpenMx's generic numeric-derivative checks return status code 6 (RED). This "threshold" is also used by the Newton-Raphson optimizer. We hope this will reduce the frequency of spurious code 6s.
  • OpenMx's nonparametric bootstrapping feature has new utility functions. For details, see documentation for mxBootstrapStdizeRAMPpaths(), omxBootstrapCov(), omxBootstrapEval(), omxBootstrapEvalCov(), and omxBootstrapEvalByName().
  • For relational-SEM models, mxGenerateData() now generates data in the compiled backend. This improves performance, and resolves a bug such that generating data for large relational models would fail.
  • OpenMx's Nelder-Mead optimizer now allows the user to request that the initial simplex be centered on the start values, instead of using them as the coordinates for a vertex.
  • The Nelder-Mead optimizer will now check to see if the final simplex's eucentroid and subcentroid have better fit values than any of its vertices; if the pseudoHessian is to be calculated, the optimizer will likewise check the midpoints of the final simplex's edges.
  • The Nelder-Mead optimizer now behaves as designed when argument ineqConstraintMthd="eqMthd" and there are no equality MxConstraints.
  • A bug in mxTryHardctsem() and mxTryHardWideSearch() has been repaired. This bug caused those functions to use a default value of 1, not 10, for argument extraTries.
  • A bug involving specification of the means vector for LISREL expectation has been repaired, and the LISREL expectation's handling of errors involving the means vector has been improved.