The AstroStat Slog

Sherpa is a fitting environment in which Chandra data (and really, X-ray data from any observatory) can be analyzed. It has just undergone a major update and now runs on python. Or allows python to run. Something like that. It is a very powerful tool, but I can never remember how to use it, and I have an amazing knack for not finding what I need in the documentation. So here is a little cheat sheet (which I will keep updating as and when if I learn more):

On Help:

• In general, to get help, use: ahelp "something" (note the quotes)
• Even more useful, type: ? wildcard to get a list of all commands that include the wildcard
• You can also do a form of autocomplete: type TAB after writing half a command to get a list of all possible completions.

Data I/O:

• To read in your PHA file, use: load_pha()
• Often for Chandra spectra, the background is included in that same file. In any case, to read it in separately, use: load_bkg()
  • Q: should it be loaded in to the same dataID as the source?
  • A: Yes.
  • A: When the background counts are present in the same file, they can be read in separately and assigned to the background via set_bkg('src',get_data('bkg')), so counts from a different file can be assigned as background to the current spectrum.
• To read in the corresponding ARF, use: load_arf()
  • Q: load_bkg_arf() for the background — should it be done before or after load_bkg(), or does it matter?
  • A: does not matter
• To read in the corresponding RMF, use: load_rmf()
  • Q: load_bkg_rmf() for the background, and same question as above
  • A: same answer as above; does not matter.
• To see the structure of the data, type: print(get_data()) and print(get_bkg())
• To select a subset of channels to analyze, use: notice_id()
• To subtract background from source data, use: subtract()
• To not subtract, to undo the subtraction, etc., use: unsubtract()
• To plot useful stuff, use: plot_data(), plot_bkg(), plot_arf(), plot_model(), plot_fit(), etc.
• (Q: how in god’s name does one avoid plotting those damned error bars? I know error bars are necessary, but when I have a spectrum with 8192 bins, I don’t want it washed out with silly 1-sigma Poisson bars. And while we are asking questions, how do I change the units on the y-axis to counts/bin? A: rumors say that plot_data(1,yerr=0) should do the trick, but it appears to be still in the development version.)

Fitting:

• To fit model to the data, command it to: fit()
• To get error bars on the fit parameters, use: projection() (or covar(), but why deliberately use a function that is guaranteed to underestimate your error bars?)
• Defining models appears to be much easier now. You can use syntax like: set_source(ModelName.ModelID+AnotherModel.ModelID2) (where you can distinguish between different instances of the same type of model using the ModelID — e.g., set_source(xsphabs.abs1*powlaw1d.psrc+powlaw1d.pbkg))
• To see what the model parameter values are, type: print(get_model())
• To change statistic, use: set_stat() (options are various chisq types, cstat, and cash)
• To change the optimization method, use: set_method() (options are levmar, moncar, neldermead, simann, simplex)