You are here

Version 2.7.11 of OpenMx officially released

OpenMx version 2.7.11 is now available through CRAN and through our virginia.edu repository. Users running R version 3.4 under Mac OS X 10.11 "El Capitan" are advised to follow our website's instructions for installing OpenMx, since the CRAN build of the package still does not work with that combination of R and MacOS versions.

New features, performance improvements, and bug-fixes in OpenMx v2.7.11 include the following:

  • OpenMx now has a nonparametric bootstrapping feature! The re-sampling and refitting of the model is carried out in the compiled "backend," which is considerably faster than doing so in interpreted R code. OpenMx can provide bootstrap standard errors, quantile confidence intervals, and bias-corrected quantile confidence intervals, both for free parameters and for functions of free parameters (e.g., elements of MxAlgebras). For details, see the documentation for mxBootstrap(), mxComputeBootstrap(), mxBootstrapEval(), and mxBootstrapEvalByName().
  • Several documentation pages have been edited for clarity.
  • Oakes standard errors (which are used with MxComputeEM compute steps) now work properly.
  • It is now possible to provide a repeated-sampling covariance matrix (of the free parameters) to mxSE() and mxStandardizeRAMpaths() (which by default calculate such a matrix from the Hessian in the model's output slot). This makes it possible to use a robust or bootstrap covariance matrix with those functions.
  • Instances where SLSQP fails to move from its start values should now be rarer.
  • Peak memory demand of the GREML fitfunction and expectation should be lower, especially in cases where the dataset contains missing values.