data.foo should throw error when 'foo' is missing

hmm... just svn'd
and did
rm(list = ls())
detach("package:OpenMx")
library(OpenMx)

and still get the same error. how odd.

Running Definition Means -- Matrix Specification
Error: The thresholds matrix/algebra 'x' for model 'Definition Means -- Matrix Specification' does not exist.

I could run it fine from the command line, and it runs fine after quitting R and reloading. So closed...

I've got a feeling that releasing the OpenMx library and reloading leaves the old version loaded and a bit unstable. Still can't understand the nature of the error.

Safest to just quit and reload in between svn updates, i guess

I see now that you can change the name of a definition variable to something that doesn't exist, and that runs fine: I think any label called "data.nnn" where nnn is not a valid column in the mxData should throw an error, shouldn't it?

model <- mxModel("Cov",
mxMatrix(type="Symm", nrow=2, ncol=2, free=TRUE, values=c(1, 0, 1), name = "Sigma"),
mxMatrix(type="Full", nrow=1, ncol=2, free=TRUE, name = "means"),
mxMatrix(type="Full", nrow=1, ncol=2, free=TRUE, values=c(0, 0), name = "beta"),
mxMatrix(type="Full", nrow=1, ncol=2, free=FALSE, labels=c("data.no_such_thing"), name = "def"),
mxAlgebra(expression=means+(beta*def), name = "Mu"),
mxFIMLObjective(covariance="Sigma", means="Mu", dimnames=selCovs),
mxData(observed=rawData, type="raw")
)

Thanks Michael ! Can I say too that the errors this generates are a model of clarity: Blind Freddy should be able to follow them (possibly an Australian-ism :-) )