Remnant-fits.md

@@ -322,7 +322,61 @@ Maximum difference for af_evol_avg: 9.770286555355057e-07
 
 * Again as a caveat, we have not looked at the special cases of spin-aligned or non-spinning PE runs yet. :construction:
 
-# surrogate, SEOBNRv4PHM, and IMRPhenomPv3HM fits
+# surrogate fits
+
+## Related MRs
+
+*  !254
+
+## example output
+
+* https://ldas-jobs.ligo.caltech.edu/~charlie.hoy/projects/pesummary_review/remnant_fits/NRSurrogate/home.html
+
+## checks against an independent script
+
+We compared against the [results](https://ldas-jobs.ligo.caltech.edu/~nathan.johnson-mcdaniel/o3-postprocessing/NRSur7dq4_nest_unreviewed_phase_fix/NRSur7dq4_nest_unreviewed_phase_fix_final_state_samples.dat) from Nathan's [independent script](https://ldas-jobs.ligo.caltech.edu/~nathan.johnson-mcdaniel/o3-postprocessing/apply_wf_fits.py) (which is wrapped with some [Condor parallelization](https://ldas-jobs.ligo.caltech.edu/~nathan.johnson-mcdaniel/o3-postprocessing/make_wf_fit_application_pipeline.py)) on [these GW190521.1 NRSur7dq4 PE samples](https://ldas-jobs.ligo.caltech.edu/~max.isi/cbc_pe/GW190521g/nr7dq4_nest_ao1_8s_nlive2048/1242442967.45-333631/V1H1L1/posplots.html).
+
+The comparison was carried out using this script
+
+```
+import numpy as np
+
+Nathan = np.genfromtxt("NRSur7dq4_nest_unreviewed_phase_fix_final_state_samples.dat", names=True)
+Charlie = np.genfromtxt("/home/charlie.hoy/public_html/projects/pesummary_review/remnant_fits/NRSurrogate/samples/NRSurrogate_pesummary.dat", names=True)
+
+Nathan_names = ["Mf_NRSur7dq4Remnant", "af_NRSur7dq4Remnant", "vf_NRSur7dq4Remnant"]
+Charlie_names = ["final_mass", "final_spin", "final_kick"]
+for lalinf, pesummary in zip(Nathan_names, Charlie_names):
+    Nathan_samps = Nathan[lalinf]
+    Charlie_samps = Charlie[pesummary]
+    difference = np.abs(Nathan_samps - Charlie_samps)
+    print("Max difference for {}: {}".format(lalinf, np.max(difference)))
+    print("Min difference for {}: {}".format(lalinf, np.min(difference)))
+```
+
+Which gave the following results:
+
+```
+Max difference for Mf_NRSur7dq4Remnant: 0.0
+Min difference for Mf_NRSur7dq4Remnant: 0.0
+Max difference for af_NRSur7dq4Remnant: 0.0
+Min difference for af_NRSur7dq4Remnant: 0.0
+Max difference for vf_NRSur7dq4Remnant: 1.3642420526593924e-12
+Min difference for vf_NRSur7dq4Remnant: 0.0
+```
+
+## Review comments
+
+* This implementation only allows the surrogate fits to be evaluated for mass ratios > 1/6, where they are expected to give good results. We have not tested what occurs if the PESummary implementation is applied to PE samples with more unequal mass ratios.
+* We have only checked this on NRSur7dq4 samples. While the logic is there for this to be applied to samples from a waveform model parameterized by the spins at `flow`, not `f_ref`, such as SEOBNRv4P[HM], we did not test that it works correctly, since the only current use is the application to NRSur7dq4 samples for GW190521.1
+
+## sign-off
+
+Nathan and David sign off on !254, as of 10th March 2023, with the reference output as here: https://ldas-jobs.ligo.caltech.edu/~charlie.hoy/projects/pesummary_review/remnant_fits/NRSurrogate/home.html ✔
+
+* Again as a caveat, we have not checked the application to SEOBNRv4P[HM] samples yet. :construction:
+
+# SEOBNRv4PHM and IMRPhenomPv3HM fits
 
 
 ## Related MRs