SEM models are becoming more difficult to specify and estimate as substantive theory and data grow increasingly complex. Massive data sets including genome–wide association and brain imaging are at the leading edge of this evolving research landscape. These data sets are many orders of magnitude larger than those available when most SEM software was originally designed and programmed. For instance, an fMRI data set might include 40,000 voxels per frame per person. An SEM model of these data might include hundreds of latent variables and tens of thousands of free parameters. In such a case, one would need a large parallel computing grid to estimate the model. OpenMx has been designed from the beginning with parallel computing in mind, both for use with multicore computers and with very large grids of computers such as the TeraGrid and Open Science Grid.
