General-Review-(post-GW190814).md

@@ -76,8 +76,11 @@ marginalizing to compute posteriors and odds ratio
 
 ## Additional checklist items for addition of spin inference
 
-  * [ ] inference reproduces the same results when spins are neglected (--spin-column is not passed) and when spins are identically zero for all samples
+  * [X] inference reproduces the same results when spins are neglected (--spin-column is not passed) and when spins are identically zero for all samples
+    - Review statement: I have checked that this is true and results can be found in the second two columns of the second table in "Comparison to Previous Results" below.
   * [ ] inference produces strictly larger estimates for P(m2<Mmax) when spins are included (but known to be less than the break-up spin) compared to when spins are neglected
+    - Review statement: I have checked this and it is true only for GW200105. Results can be found in the first two columns of the second table in "Comparison to Previous Results" below. The descrepancy between GW200105 and GW200115 is likely because of the fact that the secondary spin posteriors on GW200115 are very broad and therefore a significant fraction (~30%) of the posterior support is above the maximum allowed spin (approximately 0.7). This causes P(m_2<Mmax|data) to occasionally be lower for the case in which we are considering spins. On the other hand, secondary spin posteriors on GW200105 are narrower and peaked at lower spins (rarely above 0.7). 
+    - This should be checked by doing a run with the low-spin priors and seeing if the --spin-column results are strictly larger than the others
   * [X] scaling of Mmax(Mtov, Rtov, spin) was correctly implemented based on Eqn 18 of [Breu+Rezzolla 2016](https://academic.oup.com/mnras/article/459/1/646/2608837).
     - Review statement: I have cross checked the function `br2016_mmax_scaling()` in `bin/mmax_model_selection/utils.py` with equation 18 in the above paper. There seems to be an error. Namely, the second term in the mmax expression should have a factor of the compactness squared, but it is currently to the 1st power. 
    - Updated review statement: This has now been corrected by Reed in commit 50fe99a74e8fa3ebf183505a59af13e053afac36 of MR !9. 
@@ -88,7 +91,7 @@ marginalizing to compute posteriors and odds ratio
   * [X] the (updated) code can be run with minimal changes by cloning the repo
     - Review statement: this is true and is documented below for the [NSBH repo](https://git.ligo.org/reed.essick/o3-nsbh-mmax-model-selection). 
     - Note that the person running needs to convert all .json and .dat files to csv files by following the steps outlined in the README. 
-    - Commands run by reviewer and output are below, and the results are consistent with those contained in the [NSBH repo](https://git.ligo.org/reed.essick/o3-nsbh-mmax-model-selection) up to 10%
+    - Commands run by reviewer are below, and the results are consistent with those contained in the [NSBH repo](https://git.ligo.org/reed.essick/o3-nsbh-mmax-model-selection) up to 10%. Output, including diffs of the reviewer's `.out` files and those found in the NSBH repo, can be found on CIT in `/home/amanda.farah/projects/mmax_review/o3-nsbh-mmax-model-selection/results_review/`
 ```
 [amanda.farah@ldas-grid mmax_review]$ git clone https://git.ligo.org/reed.essick/mmax-[amanda.farah@ldas-grid mmax_review]$ model-selection
 [amanda.farah@ldas-grid mmax_review]$ git fetch
@@ -309,3 +312,13 @@ Compare to review done of GW190814, found [here](https://git.ligo.org/reed.essic
 | `etc/mmax_spec.csv` | `posterior probability: P(m<=Mmax\|data) = 3.464760e-02 +/- 1.147533e-05 +/- 3.210484e-03` | `P(m<=Mmax) = 3.464760e-02 +/- 3.212052e-03`|
 | `etc/mmax_obs.csv` | `posterior probability: P(m<=Mmax\|data) = 1.037933e-01 +/- 4.245073e-06 +/- 2.037448e-03` | `P(m<=Mmax) = 1.037933e-01 +/- 2.040315e-03`|
 | `etc/just-Mmax\=2.3.csv` | `posterior probability: P(m<=Mmax\|data) = 7.210845e-05 +/- 5.098654e-05 +/- 0.000000e+00` | `P(m<=Mmax) = 7.210845e-05 +/- 5.098654e-05`|
+
+Compare spins inference to one done without spins. Here, we just look at the O3a_ALL_BROKEN_POWERLAW mass distribution, event GW200105 analyzed with a high spin prior, and LEC samples, (label `GW200105-C01_PHMcombined_highpin_pesummary-LEC-mtov-rtov`) but other cases can be found by looking at the output files.
+
+| Spin column | posterior probability P(m<=Mmax|data) | difference from above |
+| ------ | ------ | ------ |
+| `a_2` | 8.970817e-01 +/- 3.699360e-03 | `--` |
+| none passed | 8.780918e-01 +/- 4.173085e-03 | smaller by ~1.90e-02 |
+| zeros |  8.780920e-01 +/- 4.173083e-03 | 0 |
+
+Output files containing details about the entries in the above table as well as runs with many different population models, different spin priors, and different events can be found in `/home/amanda.farah/projects/mmax_review/o3-nsbh-mmax-model-selection/results_review/`
\ No newline at end of file