Next: SMHELP*
Up: SGRID
Previous: Examples
Errors arising in SGRID fall into three catagories,
- Errors generated by spectral models due to inappropriate
parameter values.
- Errors generated during minimisation when there are free
parameters remaining after the grid axes are defined, eg.
matrix of second derivatives of chi-squared or Cash
statistic (with respect to the parameters) may become
singular or ill-conditioned.
- Failure to optimise at a grid point due to insufficient
iterations.
In the first two cases SGRID will generate a bad quality (MISSING)
grid point for the combination of grid parameters giving the error.
If this is a problem then that area of parameter space giving the
the errors should not be gridded.
The third case is handled differently. On the assumption that a
good fit would have been achieved had sufficient iterations been
allowed, SGRID uses the value of the statistic but flags it by
setting the IGNORE quality bit. The quality mask is set so that
these points will be processed by other Asterix software. If it
is required that these points be treated as bad like the first
two cases above, then this can be achieved by using the MASK
application to set the quality mask to 11111111.
Web Master
Tue Oct 7 10:18:50 BST 1997