Repeatability across platforms
Posted on
Steve
Joined: 07/30/2009
| Attachment | Size |
|---|---|
| LDEUnivariateExample091116.R | 2.89 KB |
| LDEUnivariateExample091116.txt | 2.88 KB |
Forums
I have run into an interesting problem. A model (which is admittedly of low stability) that I used in class today ran on the Mac (with warnings, but reasonably close estimates) but not on the PC (non-invertible). Both are Intel processors, both running R 2.9.2 and OpenMx 0.2.2. The model is of simulated data with boundary constraints on residuals because variance explained is very high.
I attach the script and data as a test case for the limits of similarities between architectures.
This code requires
This code requires "GLLAfunctions.R" to run.
Log in or register to post comments
from my (humble)
from my (humble) understanding of R, cross platform uniformity is not guaranteed. there can be subtle differences in C, fortran, and other compilers that can generate differences so that the compiled modules do not always give exact answers across platforms.
numerical optimization with constraints is one of those situations where slight differences in compiler settings can generate discrepant results within a platform, let alone across platforms. (most numerical guru's that i've talked to always recommend unconstrained optimization for likelihood problems). suggest trying a model without constraints and then comparing the results across platforms.
if you cannot get a model without constraints to converge, then there is probably something the matter with the model, given the data.
greg
Log in or register to post comments