Hi. I've been doing some work on a package to overlay openmx and manage continuous time SEM models. Everything is working great but I am quite stuck with what seems to be some sort of memory management issue with openmx. I've spoken with Tim Brick regarding this previously, but thought I'd put it up here for all to see :)
Basically I'm estimating n-variate vector autoregressive models, and constraining the discrete observations to an underlying continuous time model with various algebra constraints and definition variables. See http://psycnet.apa.org/index.cfm?fa=search.displayRecord&id=7833EC1B-FEBA-3F1B-2A07-5321A5AE363A&resultID=4&page=1&dbTab=all&search=true for more details.
When each individual shares the same pattern of definition variables, ie all individuals are measured at the same time for each wave, things are fine. However as soon as individuals vary in their measurement timings, memory usage skyrockets and in many cases I'm unable to complete mxRun without R memory errors.
2 data files and an openmx script, reflecting a bivariate, 5 time point case are available here:
data1 is rounded such that individuals share the same time intervals (seen in variables i1 to i4)
data2 shows individually varying time intervals
the openmx script runs very quickly when data1 is used, but crashes out on me when data2 is used.
Cheers for any help!