inspiral_lr, inspiral_extrinsics: more snr chisq tuning

gstlal-inspiral/python/stats/inspiral_extrinsics.py

@@ -945,7 +945,7 @@ class NumeratorSNRCHIPDF(rate.BinnedLnPDF):
 		snr2 = snr**2.
 		ncparam_per_pf = snr2
 		# takes into account the mean depending on noncentrality parameter
-		snrchi2 = numpy.outer(snr2 * df * (1.0 + numpy.mean(pfs)), rcoss)
+		snrchi2 = numpy.outer(snr2 * df * (1.0 + max(pfs)), rcoss)
 
 		arr = numpy.zeros_like(lnpdf.array)
 		for pf in pfs:

gstlal-inspiral/python/stats/inspiral_lr.py

@@ -423,7 +423,7 @@ class LnSignalDensity(LnLRDensity):
 		vtdict = self.horizon_history.functional_integral_dict(window.shift(float(gps)), lambda D: D**3.)
 		return dict((instrument, (vt / t)**(1./3.)) for instrument, vt in vtdict.items())
 
-	def add_signal_model(self, prefactors_range = (0.001, 0.30), df = 150, inv_snr_pow = 4.):
+	def add_signal_model(self, prefactors_range = (0.001, 0.30), df = 100, inv_snr_pow = 4.):
 		# normalize to 10 *mi*llion signals.  this count makes the
 		# density estimation code choose a suitable kernel size
 		inspiral_extrinsics.NumeratorSNRCHIPDF.add_signal_model(self.densities["snr_chi"], 1e12, prefactors_range, df, inv_snr_pow = inv_snr_pow, snr_min = self.snr_min)