Review.md

@@ -60,7 +60,7 @@ Remember that `alpha=1.0` posteriors should look bad for these events - they wou
 
 | Event | reviewer | wiki page | review status :x: / :heavy_check_mark: | review comment |
 | ------ | ------ | ------ | ------ | ------ |
-| GW191204_171526 | Krishnendu | [link](O3b-Results/GW191204_171526) |  |<br> I think we already discussed this in the past about the differences in the $`A_0`$ cases from Bilby LI runs [like this](https://git.ligo.org/tomasz.baka/liv-automation/-/wikis/uploads/793638864a52c6269d894eb00cc30882/GW191204_171526_0.png). Could you please add a note about our final understanding? I believe adding one note on this will be useful as we have similar cases for multiple events and we discussed this feature in detail. **HK:** Inconsistent results are observed for events with poor or uninformative constraints on the MDR parameters. In the lalinference runs, the sampling of the positive and negative sides of the derived parameter  $`\lambda_{eff}`$ is done separately. To obtain the final $`A_\alpha`$ posteriors, the joined posterior of $`\lambda_{eff}`$ is obtained by stitching together kernel density estimates (KDEs) constructed from the $`\lambda_{eff}`$  samples. Then,  $`A_\alpha`$ is resampled from the joined KDE. However, in the new implementation, the sampling is much more direct. This seems to be the reason for the inconsistencies in the cases where the events are uninformative. <br><br> Another thing is the slight difference in the [Mc estimate](https://ldas-jobs.ligo.caltech.edu/~haris.k/bilby_pipe/liv/live_review/summarypages/o3b/GW191204_171526/0p0/plots/combined_1d_posterior_chirp_mass.png), I guess we need to note down about this feature also as part of signing off the results. <br><br> **Krishnendu** We can briefly mention this in the review statement **TB:** I see - the linked pages are older, without all the different posteriors plotted. If you look at this [link](https://ldas-jobs.ligo.caltech.edu/~tomasz.baka/LIV_review/pesummary/GW191204/0.0/html/Comparison_chirp_mass.html), you can see that while there is a difference between LIV and bilby results, bilby-liv and bilby-gr posteriors are looking very similar. The difference must be caused by some bilby settings. I will update the links to point to more in depth histograms. **Krishnendu** @**TB**, planning to add new histograms?|
+| GW191204_171526 | Krishnendu | [link](O3b-Results/GW191204_171526) |  |<br> I think we already discussed this in the past about the differences in the $`A_0`$ cases from Bilby LI runs [like this](https://git.ligo.org/tomasz.baka/liv-automation/-/wikis/uploads/793638864a52c6269d894eb00cc30882/GW191204_171526_0.png). Could you please add a note about our final understanding? I believe adding one note on this will be useful as we have similar cases for multiple events and we discussed this feature in detail. **HK:** Inconsistent results are observed for events with poor or uninformative constraints on the MDR parameters. In the lalinference runs, the sampling of the positive and negative sides of the derived parameter  $`\lambda_{eff}`$ is done separately. To obtain the final $`A_\alpha`$ posteriors, the joined posterior of $`\lambda_{eff}`$ is obtained by stitching together kernel density estimates (KDEs) constructed from the $`\lambda_{eff}`$  samples. Then,  $`A_\alpha`$ is resampled from the joined KDE. However, in the new implementation, the sampling is much more direct. This seems to be the reason for the inconsistencies in the cases where the events are uninformative. <br><br> Another thing is the slight difference in the [Mc estimate](https://ldas-jobs.ligo.caltech.edu/~haris.k/bilby_pipe/liv/live_review/summarypages/o3b/GW191204_171526/0p0/plots/combined_1d_posterior_chirp_mass.png), I guess we need to note down about this feature also as part of signing off the results. <br>**Krishnendu** We can briefly mention this in the review statement <br>**TB:** I see - the linked pages are older, without all the different posteriors plotted. If you look at this [link](https://ldas-jobs.ligo.caltech.edu/~tomasz.baka/LIV_review/pesummary/GW191204/0.0/html/Comparison_chirp_mass.html), you can see that while there is a difference between LIV and bilby results, bilby-liv and bilby-gr posteriors are looking very similar. The difference must be caused by some bilby settings. I will update the links to point to more in depth histograms. <br>**Krishnendu** @**TB**, planning to add new histograms?|
 | GW191222A | Krishnendu | [link](O3b-Results/GW191222A) |  | <br> Shall we check if we are using the same distance priors for both LI and Bilby runs?  **TB:** we checked and we are using the same distance priors|
 | GW200219A | Krishnendu | [link](O3b-Results/GW200219A) |  | <br> $`\chi_{eff}`$ posterior is different, do we understand this?  **TB**: chi_eff posteriors for bilby-liv look to me very similar to other ones before reweighting. After reweighting it is broader, but it should not be compared with lal posterior (during o3 only 1D A_alpha posterior was reweighted)|
 | GW200202A | Krishnendu | [link](O3b-Results/GW200202A) |  | <br> $`\chi_{eff}`$ posterior is different, do we understand this? <br> <br> I think it will be also good to compare the prior distributions used for LI and Bilby runs to see these differences are caused by the difference in prior. |