Next: Input Output Data Up: Spectral Fitting commands Prev: Examples

SERROR

Evaluates a confidence region corresponding to a specified chi-squared increase, along each parameter dimension in the fit space. A chi-squared increase of unity gives approximate 1 sigma errors for each parameter individually - i.e. there is a 68% probability of a parameter lying within its confidence interval where the other parameters are free to have any value at all.

Note that the parameter space volume defined by the intersection of the 1 sigma intervals for each parameter is NOT a 1 sigma confidence volume, it is considerably smaller (i.e. it contains less than 68% of the probability).

If 90% confidence intervals are required, the corresponding chi-squared increase is 2.71.

Not all parameters need have errors calculated. The hidden PARS parameter contains the list of parameter numbers to process - see examples for usage.

For a likelihood evaluation of errors, using the Cash statistic, SERROR should be invoked with LIK on the command line.

Note that the first run of SERROR will not always produce an accurate confidence interval for every parameter - read the WARNING subsection below for further information.


Subtopics:


Next: Input Output Data Up: Spectral Fitting commands Prev: Examples