Hi @leo-singer and @michael-coughlin, can you review this MR?
VINAYA VALSAN (87fb56c6) at 27 Mar 14:40
added spelling list
VINAYA VALSAN (5e0687e5) at 27 Mar 14:20
applied suggestion
@michael-coughlin, I modified the text to make it shorter.
VINAYA VALSAN (6efa1c1e) at 27 Mar 13:19
removed some details
Hi @leo-singer and @michael-coughlin, I have updated the text.
Yes. I can do that
@leo-singer, "Classification: Four numbers, summing to unity, giving probability that the source belongs to the following four mutually exclusive categories:" this statement in the user guide is true for rapidpe. The source classification numbers that rapidpe provide agrees p_bns+p_bbh+p_nsbh+p_terr=1
. The p_terr
here comes from the search pipeline source classification.
If I notate the source probability from rapidpe with a '
p^{'}_{bns}+p^{'}_{bbh}+p^{'}_{nsbh} = 1
We need,
p_{bns}+p_{bbh}+p_{nsbh}+p_{terr} = 1
where p_{terr}
is coming from search pipelines's source classification.
To go from the first equation to the second one
p_{bns} = p^{'}_{bns}(1-p_{terr})\\
p_{bbh} = p^{'}_{bbh}(1-p_{terr})\\
p_{nsbh} = p^{'}_{nsbh}(1-p_{terr})
The numbers that rapidpe includes in the alert are p_{bns}, p_{bbh}, p_{nsbh}, p_{terr}
.
This is how it has been done on update alerts in O4a. The one difference between O4a and O4b is that in O4b we do this renormalization before uploading the pastro file to gracedb and just before writing the alert content whereas, for the updated circular during O4a the normalization is done when rapidpe uploads a pastro file to gracedb.
@leo-singer, RapidPE calculates the source probability such that p_bns+p_bbh+p_nsbh=1
. But, in the rapidpe pastro file, it will have a p_terr which is taken from the pipeline and the p_bns, p_bbh, p_nsbh gets normalized such that p_bns+p_bbh+p_nsbh=1-p_terr
. Since Rapidpe is a parameter estimation pipeline, it can not compute p_terr. But, since the source probability gets renormalized by the pipeline p_terr, effectively the RapidPE source classification does not assume that the event is purely non-terrestrial.
I have modified the wording a little bit to avoid the confusion. Let me know if this is not clear.
VINAYA VALSAN (5e2f5a4e) at 22 Mar 14:42
addressed comments
explained in this comment !253 (comment 974325)
@leo-singer, the latter is correct. Rapidpe-RIFT source classification supersedes the classification from search pipelines. But, Rapidpe does not estimate the terrestrial probability. It takes the terrestrial probability from the most recent preferred event and renormalises Rapidpe's pBBH, pBNS and pNSBH accordingly.
VINAYA VALSAN (6cfa368b) at 20 Mar 16:57
fixed citation
Hi @leo-singer , this is ready to be reviewed.
VINAYA VALSAN (e3a4b3b0) at 20 Mar 14:52
spelling fix
VINAYA VALSAN (b1da7d1e) at 20 Mar 14:40
minor changes