OpenMx Developer Forums

Hi! My favorite web-site when preparing scripts for running twin models on Mx has been the ’Mx scripts Library’ http://www.psy.vu.nl/mxbib/index.php?page=home
It would have been very helpful to have all the scripts in this library converted to OpenMx.

It was suggested at last week's developers meeting that it could be useful to use the extract operator for MxMatrix objects. So I've implemented foo[x,y] and foo[x,y] <- z where 'foo' is a MxMatrix object. The new feature has been checked into the subversion repository.

Some notes about foo[x,y]

• returns a MxMatrix object
• applies [x,y] extraction to the components of foo (values, labels, free, lbound, ubound)
• returns a MxMatrix of the same type as foo, unless that would cause an error. In cases of errors, return a Full matrix.

Here are the minutes from today's developers meeting.

1. write a different mxThresholds() conformability checking when there are multiple matrices/algebras
2. add a special case for a logical vector of NAs
3. set function precision for the ordinal case to 1e-09
4. can we add constants to constraints?
5. look at filling in the values for diagonal matrices
6. implement d2v() function
7. implement v2d() function
8. add to the summary() and output the OpenMx version number
9. write a function that prints the covariance and means matrices
   

As of r1033 in the repository, the OpenMx package has its own name space when it is loaded in R. This is the "right" way to publish packages in R, and we intended to add a name space some time ago. But the namespace mechanics are scary and poorly documented.

Now the following names are exported to the user:

• Any function of the form: mx***()
• Any function of the form: omx***()
• Matrix functions implemented in the library: tr, "%&%", vech, vechs
• S4 operators: "[","[[", "$"
• S4 methods: summary, nrow, ncol, length, dimnames, show, print

Happy new years! Here is the TODO list generated from today's developer meeting.

• Add 'hide' argument to mxRun(). Default value is TRUE. Determines whether to return the hidden matrices and algebra generated during the workflow.
• Add start and stop wall-clock times to the output of mxRun(). Separate run time into front-end time and back-end time.
• Implement \mnor() in OpenMx before the Twin Workshop.
• Fix the bugs in FIMLRowAlgebra.R example.
• Add page on the OpenMx site for error messages
• Fix the bug in categoricalDefinitionTest.R example.
  

Sometimes the models in trunk/models/passing, failing, etc. are useful to link to, even if they are not part of the official docs (yet). How does one compose a URL to get to these directories?

This proposal is for a new objective function that allows for a generic row-by-row computation on a data-set, followed by a final computation to aggregate all the row-by-row results. The objective function takes four arguments.

mxRowObjective(rowAlgebra, rowResults = NA, reduceAlgebra = NA, name = NA).

A recently posted issue requested the implementation of Mx1's \mnor and \allint functions, which perform integrations between thresholds on a multivariate normal distribution. This functionality is already being used in FIML Objectives if the data are ordinal (using code supplied by Alan Genz), so the back-end implementation is not difficult.

In r1011, some improvements have been made to definition variables. Definition variables can now be used inside algebra expressions. Additionally, definition variables inside of MxMatrices will populate to the 1st row before conformability checking. What this means in plain English: you do not need to specify the starting values for definition variables.