Changes

Page history
Update SummaryReview script authored by Nicola De Lillo's avatar Nicola De Lillo
Show whitespace changes
Inline Side-by-side
SummaryReview-script.md
View page @ e5376ff6
...@@ -10,7 +10,7 @@ Can be seen here. ...@@ -10,7 +10,7 @@ Can be seen here.
# Review comments # Review comments
#### Step 1 (Run) #### Step 1 (Run)
We run `pesummary` and `cbcBayesPostProc` on the posterior file `.hdf5` indicated at [PESummary-Review#Notes](https://git.ligo.org/lscsoft/pesummary/-/wikis/PESummary-Review#notes) I run `pesummary` and `cbcBayesPostProc` on the posterior file `.hdf5` indicated at [PESummary-Review#Notes](https://git.ligo.org/lscsoft/pesummary/-/wikis/PESummary-Review#notes)
```bash ```bash
summaryreview -w /home/nicola.delillo/public_html/pesummary-review/summaryreview -s /home/nicola.delillo/pesummary-review/samples/PROD0_posterior_samples.hdf5 summaryreview -w /home/nicola.delillo/public_html/pesummary-review/summaryreview -s /home/nicola.delillo/pesummary-review/samples/PROD0_posterior_samples.hdf5
...@@ -27,7 +27,7 @@ $ lalsim.GetApproximantFromString('IMRPhenomPv2_NRtidal') ...@@ -27,7 +27,7 @@ $ lalsim.GetApproximantFromString('IMRPhenomPv2_NRtidal')
79 79
``` ```
We edit the file running I edit the file running
```bash ```bash
f = h5py.File('PROD0_posterior_samples.hdf5', 'a') f = h5py.File('PROD0_posterior_samples.hdf5', 'a')
...@@ -48,11 +48,7 @@ The single summary pages can be found at: ...@@ -48,11 +48,7 @@ The single summary pages can be found at:
In `pesummary` the credible interval for 1D histograms is calculated in a way that includes the 90% of the probability: [median - 5% percentile, median + 95%], as you can see from [plot.py#L106](https://git.ligo.org/lscsoft/pesummary/-/blob/master/pesummary/core/plots/plot.py#L106) and [plot.py#L107](https://git.ligo.org/lscsoft/pesummary/-/blob/master/pesummary/core/plots/plot.py#L107) In `pesummary` the credible interval for 1D histograms is calculated in a way that includes the 90% of the probability: [median - 5% percentile, median + 95%], as you can see from [plot.py#L106](https://git.ligo.org/lscsoft/pesummary/-/blob/master/pesummary/core/plots/plot.py#L106) and [plot.py#L107](https://git.ligo.org/lscsoft/pesummary/-/blob/master/pesummary/core/plots/plot.py#L107)
in the source code. in the source code.
To review the credible interval produced, we use To review the credible interval produced, have run this script that computes and prints the credible intervals as output of pesummary attribute `.confidence_interval`
We have run this script that computes prints the credible intervals as output
of pesummary attribute `.confidence_interval`
```python ```python
...@@ -116,13 +112,61 @@ L1_matched_filter_snr_angle [-0.07 0.06] ...@@ -116,13 +112,61 @@ L1_matched_filter_snr_angle [-0.07 0.06]
a_1 [0. 0.11] a_1 [0. 0.11]
``` ```
These numbers have been compared to the reviewed results for GW190814 `combinedPHM` at the bottom of this page: [https://git.ligo.org/publications/gw190814/gw190814-discovery/-/wikis/PE-result-table](https://git.ligo.org/publications/gw190814/gw190814-discovery/-/wikis/PE-result-table) These numbers have been compared to the reviewed results for GW190814 `combinedPHM` at the bottom of this page [1]
* upper value `mass ratio`
* lower value `total_mass_source`
* lower value `mass_1_source`
* lower limit `theta_jn `
* both limits for `L1_matched_filter_snr_abs ` (0.05 difference for the lower limit)
* Same for `H1_matched_filter_snr_abs `
* Same for `V1_matched_filter_snr_abs` , with a 0.13 difference.
My guess is that this is due to different versions of `np.percentile` (pesummary uses ). To cross-check this, I have personally run the script at the page linked [1]:
```
import numpy as np
from pesummary.gw.file.read import read
parameter_dict = {
"chirp_mass_source": "Chirp mass $\mathcal{M}/M_{\odot}$",
"total_mass_source": "Total mass $M/M_{\odot}$",
"mass_ratio": "Mass ratio $q$",
"mass_1_source": "Primary mass $m_{1}/M_{\odot}$",
"mass_2_source": "Secondary mass ${m_{2}/M_{\odot}}$",
"chi_eff": "Effective inspiral spin parameter $\chi_{eff}$",
"chi_p": "Effective precession parameter $\chi_{p}$",
"a_1": "Dimensionless primary spin magnitude $a_{1}$",
"luminosity_distance": "Luminosity Distance $D_{L}/\text{Mpc}$",
"redshift": "Source redshift $z$",
"theta_jn": "Inclination angle $\\theta_{\mathrm{JN}}$",
"L1_matched_filter_snr": "Signal to Noise ratio in \\llo $\\rho_{\mathrm{L}}$",
"H1_matched_filter_snr": "Signal to Noise ratio in \\lho $\\rho_{\mathrm{H}}$",
"V1_matched_filter_snr": "Signal to Noise ratio in \\virgo $\\rho_{\mathrm{V}}$",
"network_matched_filter_snr": "Network Signal to Noise ratio $\\rho_{\mathrm{HLV}}$",
}
interested = ["PhenomPv3HM", "SEOBNRv4PHM", "combinedPHM"]
for waveform_model in interested:
print("---------- {} ----------\n".format(waveform_model))
f = read(result_file_map[waveform_model])
f.generate_all_posterior_samples()
params = parameter_dict.keys()
for param in params:
samples = f.samples_dict[param]
print(
"{} = {}^{}_{}".format(
param, np.round(np.median(samples), 2),
np.round(np.percentile(samples, 95) - np.median(samples), 2),
np.round(np.median(samples) - np.percentile(samples, 5), 2)
)
)
```
* Check upper value `mass ratio` [1]: [https://git.ligo.org/publications/gw190814/gw190814-discovery/-/wikis/PE-result-table](https://git.ligo.org/publications/gw190814/gw190814-discovery/-/wikis/PE-result-table) . There are sum difference in the following parameters
* Check lower value `total_mass_source` \ No newline at end of file
* Check lower value `mass_1_source`
* Check lower limit `theta_jn `
* Check both limits for `L1_matched_filter_snr_abs ` (0.05 difference for the lower limit)
* Same issue with `H1_matched_filter_snr_abs `
* Same issue with `V1_matched_filter_snr_abs` , with a 0.13 difference.