Loving ISIS - Confessions of a Former XSPEC User

Expand your mind! The image on the left was made with the S-lang HDF5 module and the volview image visualizer, all run with ISIS.

Making the Leap

ISIS started out as a Chandra gratings/wavelength-centric program; however, it evolved to become much more than that. This heritage as a Chandra gratings analysis tool shows up in some ways, while its design philosophy as a powerfully scripted program and the fact that all ISIS commands are S-lang functions show up in other ways. These difference take a little getting used to when coming from the XSPEC world:

• All ISIS commands are S-lang functions, and many have data return values. These return values can be captured in variables and used in interactive analysis or scripts, or they can be ignored.
• The ISIS defaults are for almost all functions, arrays, etc., to be in wavelength ascending order.
• ISIS presumes that many things, such as grouping data, will be done during analysis, and therefore it will ignore default binning in FITS file headers.
• All ISIS commands end with a semi-colon (;).

These latter two default behaviors can be overridden via choices made in a standard start-up file. Specifically, create in your home directory a file called .isisrc and add the lines:

   Isis_Append_Semicolon=1;
   Isis_Use_PHA_Grouping=1;

Example Analysis

To help familiarize users with these differences, on these web pages I give a detailed example of an ISIS analysis of a set of simultaneous radio and RXTE X-ray observations. The former is stored as an ASCII file giving frequency in Hz and radio flux (with its, presumed gaussian, error) in units of mJy. The latter consists of both PCA and HEXTE data, stored as FITS files, taken from the usual RXTE data extraction procedures. The data can be obtained as a gzipped tar file found here.

A plot of the data is shown below. A detailed description of the observations can be found in Nowak et al. (2005), ApJ, 626, p. 1006. (Every figure and result in that paper was generated with ISIS.)

All spectral fit results and figures found on these web pages were generated with this analysis script. This script presumes that you have downloaded my .isisrc file. On my 2.8 GHz Intel Core 2 Duo MacBook Pro, the whole script, soup to nuts (four different models, two complete sets of error bars, and a 961 point grid of error contours), runs in about a half hour. Whereas this might seem like a long time, that's mostly driven by the addition of the radio data points (sitting 8 orders of magnitude away in energy space makes for a slightly statistically odd lever arm; there can be some "fussiness" with the fitting the location of the radio/X-ray spectral break), and the presence of the computationally expensive reflection and diskline models. If one were to restrict this to only X-ray data being fit with broken power laws and gaussians, the whole script would run in minutes. A shortened analysis, which simultaneously fits the radio but only considers broken power law plus gaussian line models, is presented on this page, and takes only 5 minutes to produce a set of fits, complete error bars, and an error contour grid.

Next up: Five ISIS Basics.