NAME

fitpars -- Parameter set for specifying the fit to be performed.

USAGE

fitpars

DESCRIPTION

This is a parameter set for specifying the attributes of a fit to be performed. The parameters include the choice of fitting algorithm, the global fit tolerance that must be met for convergence, and the maximum number of iterations; they also include a choice of methods to estimate the parameter uncertainties.

Note that this is a pset, not an executable task; it defines a set of parameters used by other tasks. Invoking the pset by name runs `eparam' on the parameter set, allowing the user to modify the parameters. Alternatively, the parameters may be modified on the command line by specifying the pset name and parameter name. For example, the user can type `fitpars.algorithm="marquadt"' to set the name of the fitting algorithm to the string "marquadt". Parameters can also be edited by using `eparam' on the calling task (e.g., by typing "epar autospec"), in which case `fitpars' will appear as one of the task parameters; the `fitpars' parameters may then be edited by positioning the cursor on the line containing the pset name and typing ":e". After editing the pset parameters, exit normally from eparam (with an EOF, usually <cntr-Z>) to return to the main task parameter menu.

More detailed help may be accessed through the "Help" button in the interactive "vuespec" task. Alternatively, you can direct your Web browser to URL http://ra.stsci.edu/ASpec.html.

PARAMETERS

(algorithm = "simplex") [string]

Choice of algorithm for the fit. The possible choices are "simplex" and "marquadt".

(fit_toler = 1.e-4) [real]

Global tolerance to define convergence for fits. The fit is said to have converged when the merit function changes by less than this fractional amount.

(max_iter = 300) [int]

Maximum number of iterations of the fit algorithm before stopping; if this limit is reached, a warning message is issued; the components are written to the output DataBase in any case.

(err_method = "curv_matrix") [string]

Choice of method for estimating the parameter uncertainties for a fit. The possible choices are "curv_matrix" and "bootstrap". NOTE: the bootstrap method has not yet been implemented! The curv_matrix method estimates the parameter errors using chi-square and the curvature matrix, while bootstrap estimates the uncertainties by modelling the change in parameter values as the the input spectra varied. For details, type "help aspec opt=sys".

(fit_domains = "*") [string]

Specification of the dispersion domain for the fit. The domain(s) may be specified like an IRAF range, or more than one domain may be specified by putting them in an ASCII file, and providing here the name of the file preceeded by the "@" sign.

REFERENCES

This package was written by J.D. Eisenhammer, S.J. Hulbert, Z.G. Levay, and R.A. Shaw. Support for this software development was provided by the Astrophysics Data Program through NASA grant NAG5-31859 to STScI.

SEE ALSO

autospec, vuespec