# New Features for Mx 1.20

####
Debugging of matrix formula errors

Improved debugging of matrix formulae. Adjacent binary operators trapped.
Syntax checking of all matrix formulae on input (rather than at function
evaluation as was previously the case for non-calculation groups). Dumping
of parsed formula enabled for frontend.

#### Global variables

Definition of global variables. Syntax:

#define nvars 3

will ascribe 3 to nvars, which can then be used in place of 3 in matrix
dimensions, parameter specifications, and other places where numbers are read
from the script. This makes it easier to change long scripts.

#### Revision of matrix formulae in multiple fit mode

Compute, covariance, m2, m1, mean, & threshold commands may be used to revise
the matrix formulae defined within a group. The group to be revised is
indicated by a (required) #set group n command prior to the new formula
definition. n must be the number of a valid group.

#### Penalty function for multinormal integral

The multinormal integral function now uses a penalty function for
non-positive
definite input matrices. It will return densities >1 for such matrices. If
the user is calculating chi-squared or some such statistics from the result of
this integration, this should steer optimization away from non-positive
definiteness. Specifically, the integral returned is 1+abs(largest negative
eigenvalue). It doesn't work very well but I'm still working on it.

#### Rectangular file read

Input of rectangular files with dots for missing values. SAS and SPSS
both
output data like this more or less by default. You can read in such files
instead of VL data by using

rectangular file=filename.rec

**Mike Neale,
***neale@gems.vcu.edu*
Medical College of Virginia