add median_sigma to fit_by_template and fit_over_multiparam (gwastro#…

…2843)

* add median_sigma to fit_by_template and fit_over_multiparam

* fix: accidentally hardcoded H1

* logging

* couple of fixes to help/comments.

prefer not to build statements that something is 'recent' into code that might stay around for years

bin/hdfcoinc/pycbc_fit_sngls_by_template

@@ -50,6 +50,7 @@ def get_stat(statchoice, trigs, threshold):
     # Return boolean area that selects the triggers above threshold
     # along with the stat values above threshold
     return np.concatenate(select), np.concatenate(stat)
+
 #### MAIN ####
 
 parser = argparse.ArgumentParser(usage="",
@@ -309,6 +310,10 @@ for tnum in trange:
     if (tnum % 1000 == 0): logging.info('Fitted template %i / %i' %
                                                     (tnum - tmin, tmax - tmin))
 
+logging.info("Calculating median sigma for each template")
+median_sigma = [np.sqrt(np.median(trigf[args.ifo + '/sigmasq'][reg])) for reg in
+                    trigf[args.ifo + '/sigmasq_template'][:]]
+
 outfile = h5py.File(args.output, 'w')
 # store template-dependent fit output
 outfile.create_dataset("template_id", data=trange)
@@ -317,6 +322,7 @@ outfile.create_dataset("fit_coeff", data=fits)
 if args.save_trig_param:
     outfile.create_dataset("template_param", data=tpars)
 outfile.create_dataset("count_in_template", data=counts_total)
+outfile.create_dataset("median_sigma", data=median_sigma)
 # add some metadata
 outfile.attrs.create("ifo", data=args.ifo.encode())
 outfile.attrs.create("fit_function", data=args.fit_function.encode())

bin/hdfcoinc/pycbc_fit_sngls_over_multiparam

@@ -44,9 +44,8 @@ parser.add_argument("--fit-param", nargs='+',
                          "template fit file or choose from mchirp, mtotal, "
                          "chi_eff, eta, tau_0, tau_3, template_duration, "
                          "a frequency cutoff in pnutils or a frequency function"
-                         "in LALSimulation. If the background is regressed over "
-                         "multiple parameters, one can simply provide them as "
-                         "a list.")
+                         "in LALSimulation. To regress the background over "
+                         "multiple parameters, provide them as a list.")
 parser.add_argument("--approximant", default="SEOBNRv4",
                     help="Approximant for template duration. Default SEOBNRv4")
 parser.add_argument("--f-lower", type=float, default=0.,
@@ -60,8 +59,7 @@ parser.add_argument("--log-param", nargs='+',
                     help="Take the log of the fit param before smoothing.")
 parser.add_argument("--smoothing-width", type=float, nargs='+', required=True,
                     help="Distance in the space of fit param values (or the "
-                         "logs of them) to smooth over. Required. For log "
-                         "template duration, try 0.2"
+                         "logs of them) to smooth over. Required. "
                          "This must be a list corresponding to the smoothing "
                          "parameters.")
 args = parser.parse_args()
@@ -95,7 +93,7 @@ for param, slog in zip(args.fit_param, args.log_param):
     else:
         raise ValueError("invalid log param argument, use 'true', or 'false'")
 
-if 'count_in_template' in fits.keys():  # recently introduced dataset
+if 'count_in_template' in fits.keys(): # older files may not have this dataset
     tcount = True
 else:
     tcount = False
@@ -132,6 +130,10 @@ outfile = h5py.File(args.output, 'w')
 outfile['template_id'] = tid
 outfile['count_above_thresh'] = nabove_smoothed
 outfile['fit_coeff'] = alpha_smoothed
+try:
+    outfile['median_sigma'] = fits['median_sigma'][:]
+except KeyError:
+    logging.info('Median_sigma dataset not present in input file')
 if tcount: outfile['count_in_template'] = ntotal_smoothed
 
 for param, vals, slog in zip(args.fit_param, parvals, args.log_param):