Loving ISIS - Confessions of a Former XSPEC User



Data backgrounds are one of those tricky areas of X-ray astronomy. Basically, a background is any part of your data that isn't described by the source of interest to you. It could be an unresolved source in your field of view, hence it represents flux that has passed through all parts of the satellite system. Or it could be something instrumental, e.g., particle hits on the detector, and therefore something that has only passed through some parts of the system. It could be something that is measured at the time of your observation, estimated from on board housekeeping telemetry, or cobbled together from archival observations of blank sky fields, etc. As I said, it's tricky. There's all sorts of ways one may or may not want to deal with it.

First let's start off by saying what ISIS doesn't do. ISIS never "subtracts" a background from the data. Total observed data counts are always fit as the sum of a spectral model folded through the detector responses, plus an estimated background. Depending upon the choice of statistic and the definition of the background data, ISIS sometimes increases the error bars on the data and in calculations of the fit statistic to account for the presence of a (measured and/or modeled) background. This behavior is essentially identical to that of XSPEC.

This is to be contrasted with the behavior of the plotting routines. These do allow the data to be plotted with the "background subtracted". Data error bars are scaled to account for the background uncertainty, and the residuals shown in the plots hue to the ISIS internal calculations of the fit statistic. But again, data to model comparisons are always done as the sum of model and background to the total data, relative to an error bar defined by the users choice of fit statistic.

In this example, we will be use all the default behaviors of ISIS as regards background. However, there are a number of ISIS functions that allow further control of any background. We list some of these here:

     get_back(id)                           % Get the background data for a dataset
     define_back(id, "file");               % Replace a background with data from a file
     _define_back(id, bgd, area, exposure)  % Replace a background with data from variables
     back_fun(id,"function")                % Use a function to directly fit the background,
                                            % i.e., without folding through the response
Furthermore, I have created two functions that make use of the back_fun function:
    corback(id); % Allow rescaling of a background by a fit constant during fitting
    corfile(id1,id2,scale); % Include a scalable background-subtracted spectrum as an "additional" background
I have used the latter function, for example, to remove "galactic ridge emission" from RXTE observations of faint galactic sources. The galactic ridge emission data is taken from another RXTE observation from when the source in question is completely quiescent. The latter function could also be used for fitting "difference spectra" (which is popular in some parts these days).

The above routines are not used in the example presented here, but I list them for completeness, and they are available in my .isisrc scripts.

Next up: Model Grids.

This page was last updated Oct 7, 2013 by Michael Nowak. To comment on it or the material presented here, send email to mnowak@space.mit.edu.
Valid HTML 4.01! Made with JED. Viewable With Any Browser.