Difference between revisions of "Analysis BLLac Aug 2021 Aguasca-Cabot"

From my_wiki
Jump to: navigation, search
Line 29: Line 29:
 
* Increase of events in each iteration: 1000
 
* Increase of events in each iteration: 1000
 
* Minimum threshold to consider two consecutive observations (min): 10
 
* Minimum threshold to consider two consecutive observations (min): 10
 +
* Reco. energy axis: [0.1,1,3], log. interp.
 +
* True energy axis: Same as high-level data reduction specifications (see [[#Data reduction specifications|Data reduction specifications]]).
  
The best spectral model obtained fitting all data:
+
The best spectral model obtained fitting all data (using parameters in [[#Data reduction specifications|data reduction specifications]]):
  
 
[[File:Spec-model-20210808-fast.png|400px]]
 
[[File:Spec-model-20210808-fast.png|400px]]
  
This model is used to compute the fluxes using forward folding
+
This model is used to compute the fluxes using forward folding. '''no fitting is performed within the algorithm'''
  
 
== Data reduction specifications ==
 
== Data reduction specifications ==
Line 41: Line 43:
 
<li style="display: inline-block;"> [[File:BLLac-high-level-specifications.png|thumb|none|700px|High-level specifications]] </li>
 
<li style="display: inline-block;"> [[File:BLLac-high-level-specifications.png|thumb|none|700px|High-level specifications]] </li>
 
</ul></div>
 
</ul></div>
 +
Data reduction specifications for the fitting spectral model of BL Lac.
 
== Light curve ==
 
== Light curve ==
  
Line 62: Line 65:
  
 
The p-value against the null hypothesis of a constant source is 1.7350606859473474e-101.
 
The p-value against the null hypothesis of a constant source is 1.7350606859473474e-101.
 +
 +
== Checks ==
 +
 +
Most of flux time intervals are lower than 1 min. We plot the flux for 1 min time interval. We check how 1-min flux are distributed compared to the algorithm fluxes.
 +
<div><ul>
 +
<li style="display: inline-block;"> [[File:Light-curve-BLLac-fast-TRETS-loosecuts-1minSubruns.png|thumb|none|600px|Fast analysis light curve with 1 min subrun flux]] </li>
 +
</ul></div>
 +
 +
Check if the weighted average flux is compatible with respect to the 3-min (left) and run-wise fluxes (center). Also, it is checked if the weighted average 3-min fluxes are compatible with respect to run-wise fluxes (right).
 +
<div><ul>
 +
<li style="display: inline-block;"> [[File:Weighted av TRETS to subruns-3min.png|thumb|none|600px|Weighted average of the fluxes obtained with the algorithm compared with the 3 min flux.]] </li>
 +
<li style="display: inline-block;"> [[File:Weighted av TRETS to runs.png|thumb|none|600px|Weighted average of the fluxes obtained with the algorithm compared with the run-wise flux.]] </li>
 +
<li style="display: inline-block;"> [[File:Weighted av subruns-3min to runs.png|thumb|none|600px|Weighted average of the 3 min fluxes compared with the run-wise flux.]] </li>
 +
</ul></div>
 +
 +
Some of the weighted average fluxes are miss-aligned with respect to the reference one because I consider only flux points within the time interval. In instance, fluxes using events from two different observations are not considere, or between two reference time bins.

Revision as of 19:22, 10 April 2022

Basis

The basis of this algorithm is explained in the following LST-Reco calls:

  • Content: Description of the algorithm and proof of concept with Crab data and simulations of different temporal profiles.
- https://indico.cta-observatory.org/event/3571/contributions/29877/attachments/19821/27495/Aguasca_Cabot_12-07-21.pdf
  • Content: Upgrade of the algorithm and application to simulations of an orbital period of HESS J0632+057.
- https://indico.cta-observatory.org/event/3659/contributions/30756/attachments/20114/27954/20211104_Aguasca_Cabot.pdf

(Gabriel Emery developed a similar algorithm in his PhD, see https://tel.archives-ouvertes.fr/tel-02983041/document)

Objectives

  • Predict what is the observation time that LST-1 would require to achieve a given significance of a transient event to plan future observations of similar transient sources. A spectral index of the source is assumed.
  • Study the variable emission of a source using the minimum time interval LST can probe this emission considering a threshold value in its significance. Again, this is done assuming a given spectral index.
  • Obtain a light curve specifying a given threshold in the significance of the source, instead of specifying a time interval.


BL Lac light curve

- Use the same DL3 data of BL Lac obtained by the BL Lac team in order to use their light curve as the reference one. Apply the algorithm to BL Lac to obtain the minimum time interval for each flux.

Fast analysis

This fast analysis aims to show preliminary results of the algorithm to BL Lac data. MC gammas are optimized to BL Lac (see data reduction specifications). The results should improve with a refined data reduction.

Parameters values of the algorithm:

  • Range of energy (TeV): 0.1 to 1
  • Threshold in Li & Ma TS: 9
  • Threshold for UL (sigma): 2
  • Increase of events in each iteration: 1000
  • Minimum threshold to consider two consecutive observations (min): 10
  • Reco. energy axis: [0.1,1,3], log. interp.
  • True energy axis: Same as high-level data reduction specifications (see Data reduction specifications).

The best spectral model obtained fitting all data (using parameters in data reduction specifications):

Spec-model-20210808-fast.png

This model is used to compute the fluxes using forward folding. no fitting is performed within the algorithm

Data reduction specifications

  • DL1 to DL3 specifications
  • High-level specifications

Data reduction specifications for the fitting spectral model of BL Lac.

Light curve

  • Fast analysis light curve.
  • Light curve from BL Lac team

We can see that the flux points produced by the algorithm are well distributed around the run-wise fluxes and around fluxes for a time interval of more or less three minutes. Also, we can see that when the 3 min fluxes increase or decrease, the 3 sigma fluxes also show an increase or decrease. For example, see fluxes at t=5.9435e4+0.05 MJD.

Notice that the light curves in the left and right figures are slightly different because we used a different data reduction and assumed spectral index.

TS detection and time interval of the fluxes

  • TS (Li & Ma) for each time interval where a flux is plotted.
  • Histogram of flux time intervals.

In the left figure, some TS values are much higher than the threshold value (9) because I used an increase of 1000 events in each iteration to reduce the computation time.

In the right figure, there are high values of time intervals because events from two different runs are used to compute the flux.

The p-value against the null hypothesis of a constant source is 1.7350606859473474e-101.

Checks

Most of flux time intervals are lower than 1 min. We plot the flux for 1 min time interval. We check how 1-min flux are distributed compared to the algorithm fluxes.

  • Fast analysis light curve with 1 min subrun flux

Check if the weighted average flux is compatible with respect to the 3-min (left) and run-wise fluxes (center). Also, it is checked if the weighted average 3-min fluxes are compatible with respect to run-wise fluxes (right).

  • Weighted average of the fluxes obtained with the algorithm compared with the 3 min flux.
  • Weighted average of the fluxes obtained with the algorithm compared with the run-wise flux.
  • Weighted average of the 3 min fluxes compared with the run-wise flux.

Some of the weighted average fluxes are miss-aligned with respect to the reference one because I consider only flux points within the time interval. In instance, fluxes using events from two different observations are not considere, or between two reference time bins.