Acausal signal study - Githubissues

caredg commented 3 years ago

Study the simulated LW samples to:

define an object cut (perhaps on photons (preferred) or electrons)
identify the variables that could be used to single out or identify the signal, and cut off the background without affecting the signal

JonaJJSJ-crypto commented 3 years ago

I our signal, it happens that 2 in 10000 events have neither reco-electrons nor reco-photons.

JonaJJSJ-crypto commented 3 years ago

I have defined cuts in pt for our signal with the following parameters:

We need energetic electrons/photons to reconstruct our secondary vertex.
We need that there is an appreciable transversal displacement to overcome the resolution' limit of the detectors. I have implemented cuts in transversal momentum at 10,15,20 GeV for either electrons or photons. In the graphs there are no distinction between Passed Events. In case that a event has neither a good electron nor a good photon there is a distinction by label in the x axis, where bad means enough pt and not means no reconstructed particles:

Bad electrons, bad photons
Either a good electron or photon
Bad electrons, not photons
Not electrons, bad photons
Not electrons, not photons Based in this results I suggest that a pt>10GeV is a good cut

caredg commented 3 years ago

@JonaJJSJ-crypto I think most of the last plots in #32 should belong here, not in the dataset issue.

JonaJJSJ-crypto commented 3 years ago

There is a trouble with the Store of mother pdg's. Previously i was not storing 556 mom pdg but this generate incomplete Ntuples. Now i have change the code so its store 556 mother pdg as 0. This is now updated in the Gen Particle analyzer for future runs. This, however, should not be a limitation since the 556 tree is stored in GenDau instead of GenPart in case its needed to analyze the particles by mother.

JonaJJSJ-crypto commented 3 years ago

Right now I'm studying how cuts in pt could affect our data. Starting by considering the distribution for our signal electrons meaning that this electrons mother is a LW electron. So far we can see that there are few events at low energy. For comparison I have also added a non-signal(not filtered by mother) gen-electrons distribution. Where we can see that the high energy contributions comes from the signal

Following I have studied how many electrons from the signal would be affected for different cuts in pt(5,10,15). As you can see there are some events in which one electrons from the signal (not both) have pt <5. In the case of pt<10 electrons very few events contain electrons with low pt and more events have only one electron with low pt. The same goes for pt<15.

JonaJJSJ-crypto commented 3 years ago

Here also I have plotted the signal as well as the non-signal electrons distribution for different cuts in pt>5,10GeV. As can be seen, in the case of pt>5 there are less than 50 electrons from the signal cut off from the distribution and we already know that these do not correspond to the same event. In the case of the non signal events we have reduce its size to over 50% of its size. In the case of pt>10, there are a few hundreds electrons cut off from the signal here we know that except for 2 events the missing electrons are not from the same event. For non-signal electrons the size reduction is not as important as in the previous case. In the case of pt>15GeV, a couple more hundreds from signal electrons are missing yet most of them do not correspond to couples of the same event, and for non signal events the distribution resembles more to the signal showing that this region is dominated by the signal.

JonaJJSJ-crypto commented 3 years ago

For completeness of this analysis, here are presented the eta distribution for signal and non-signal generated electrons. We can see that these distributions has an expected behavior. For cuts as low as 5GeV there non-signal distribution resembles the signal distribution.

JonaJJSJ-crypto commented 3 years ago

To continue this study I have study the number of events that contains Electrons with a given cut in pt. In this case, there are 4 possible states for each event:

if neither electrons nor photons fulfill the required cut
if either electrons or photons fulfilling the required cut.
if there are not electrons fulfilling the required cut and there is no reconstructed photons.
if there are not reconstructed electrons and there is not photons fulfilling the cut.
if there is neither reconstructed electrons no reconstructed photons. From this distribution we can say that there are some events in the state 4, Its been studied and this because this events the Electrons from lw decay are pointing both at pretty close to -3, in one case, and in the other case the electrons are pointing at eta pretty close to 0.

JonaJJSJ-crypto commented 3 years ago

Here are presented the Distribution in pt and eta for each cut similar to gen particles, but something important to say is that in the case of reconstructed electrons we can se that the eta distribution is not fully reconstructed, this can be seen in the peak present in the pseudorapidity. This is not a problem since this anomaly disappear with cuts at low energy which is to be expected.

JonaJJSJ-crypto commented 3 years ago

With respect to electrons, all the study here presented suggest that a safe cut to store reconstructed electrons is 5GeV. This is because this cut in pt do not affect the signal in a significant way, and also we may avoid anomalies as those presented in this analysis. A similar cut for reconstructed electrons had been seen in other analysis with open data. This could be done to avoid anomalies, but this could be also just to reduce the size of the output files. In the case of this analysis this is a good cut and this will be implemented in the analyzer. If is the case of the filter per event I will suggest that if we require and event to have at least one energetic electrons is sufficient enough, this meaning that the filter will require that the event has at least one electron with pt>10GeV.

JonaJJSJ-crypto commented 3 years ago

Following a similar line of analysis to electrons I have studied photons in the event, because its not needed to specify the steps I will not detail each step in this part, yet if any comments is necessary it will be mentioned. There is a major difference with respect to the analysis in electrons, this is because there is no reconstructed photons with pt lower than 7GeV. This will set the cuts in pt for this analysis to pt>10,12,15,17GeV

JonaJJSJ-crypto commented 3 years ago

Generated photons: There is no evidence that the distribution of photons is affected by that the signal, because there are no important contributions in the region. From here we can see that the distribution with out cuts are hardly dominated by low energetic photons, even a cut as low as 10 GeV represents a big difference in the number of photons

JonaJJSJ-crypto commented 3 years ago

Reconstructed photons: Here we can see an anomaly in the eta distribution similar to the one seen in electrons. This anomaly also vanishes when cutting photons pt at 15GeV. Another important result is that Low energetic photons are pretty numerous at any level and they do not represent any contribution to our study. For which I suggest to cut reconstructed photons in pt >15GeV. This is still preserves a region outside the region of our signal.

caredg commented 3 years ago

@JonaJJSJ-crypto, can you check that, instead of applying a cut on the pt of the electrons, if you apply a cut on the pt of the tracks instead of the pT of the electrons, do you see the same cleaning effect? Could you check pT 7, 10, 15 and see if the peak on the pT of the electrons disappear?

caredg commented 3 years ago

@JonaJJSJ-crypto could you check the distributions, for gen and reco, of the hardest and second hardest electon (in pT) in the events? Then you impose the same set cuts for low pT on the event (but just for electron) and check these distributions again.

caredg commented 3 years ago

@JonaJJSJ-crypto, when you have this figured out, can you check on the expected size reduction for data and signal files? This should be done for:

cut on pt for at least one electron
cut on pt for at least one track
cut on pt for at least two electron
cut on pt for at least two tracks

JonaJJSJ-crypto commented 3 years ago

Here ill present the most energetic pairs of electrons for reconstructed electrons, this is because I am assuming that signal Generated electrons are the most energetic in pt. Here can be seen that this distributions resembles those of the signal

caredg commented 3 years ago

@JonaJJSJ-crypto I think we are more interested in seeing the individual distributions, right? I.e., the pT distribution for the highest-pT reconstructed electron in the event for all events, and the likewise distribution for the second highest-pT reconstructed electron in the event for all events, applying no cut and the "at least one electron with more than" 5, 10 and 15 GeV cut. The idea is that that will give us certain intuition on the kinematics of the electrons from LW particles.

JonaJJSJ-crypto commented 3 years ago

Before doing that analysis I will include the distribution for tracks, for a later comparison. This distributions don't say much but here has not been consider restrictions on chi2 for tracks. Important remarks:

there is a hard cut in pseudorapidty
there are anomalies as those seen in electrons and photons, but these do not vanish for cuts there are not presented but this had been tested for pt>20GeV also
There is no way to know how much reduces the density of tracks per events from no restrictions to 5GeV, yet we can see that the size from 5GeV to 15GeV cut the stored tracks reduce almost a 50%.

caredg commented 3 years ago

@JonaJJSJ-crypto some comments on your comments (in the same order):

yes, this seems expected, see #38
I am not sure if these are anomalies, they look more like fluctuations, statistically much less significant than that appearing in electron plots
sure, there is a decrease in the average number of tracks per event with a cut applied.

@JonaJJSJ-crypto can you check if the electron_eta and phi anomaly disappear if you impose a cut of "only events with at least one track with pT>10,15" in the event?

JonaJJSJ-crypto commented 3 years ago

I have plotted eta vs pt vs chi2, it seems that these anomalies are mostly bunches of tracks grouped in pt and eta, that im assuming are jets. and other that are grouped in chi2 and eta, but im not sure what those are

JonaJJSJ-crypto commented 3 years ago

@JonaJJSJ-crypto some comments on your comments (in the same order):
* yes, this seems expected, see [Check eta restricted region in another MC sample #38](https://github.com/JonaJJSJ-crypto/Proyecto-de-Tesis/issues/38)

* I am not sure if these are anomalies, they look more like fluctuations, statistically much less significant than that appearing in electron plots

* sure, there is a decrease in the average number of tracks per event with a cut applied.
@JonaJJSJ-crypto can you check if the electron_eta and phi anomaly disappear if you impose a cut of "only events with at least one track with pT>10,15" in the event?

These anomalies does not disappear by requiring the presence of energetic electrons. This study is similar to that which can be found in the Trigger study #32.

caredg commented 3 years ago

These anomalies does not disappear by requiring the presence of energetic electrons. This study is similar to that which can be found in the Trigger study #32.

@JonaJJSJ-crypto, that's not what I mean. Do not require anything on electons, but plot the reco electrons phi and eta after requiring the cut on the energetic track.

JonaJJSJ-crypto commented 3 years ago

@JonaJJSJ-crypto I think we are more interested in seeing the individual distributions, right? I.e., the pT distribution for the highest-pT reconstructed electron in the event for all events, and the likewise distribution for the second highest-pT reconstructed electron in the event for all events, applying no cut and the "at least one electron with more than" 5, 10 and 15 GeV cut. The idea is that that will give us certain intuition on the kinematics of the electrons from LW particles.

Actually this makes much more sense in the study of how to filter the events. Here are presented the new histograms. From this study we can see that there are events that contain 0 electrons I think this is because they are pointing towards the ends of the barrel. For those events that contain energetic electrons cuts in pt does not represent a major change in size but there is a change in size. In the case of tracks its notorious that there the last cut in pt>15GeV, and just this cut, removed a few events.

JonaJJSJ-crypto commented 3 years ago

@JonaJJSJ-crypto some comments on your comments (in the same order):
* yes, this seems expected, see [Check eta restricted region in another MC sample #38](https://github.com/JonaJJSJ-crypto/Proyecto-de-Tesis/issues/38)

* I am not sure if these are anomalies, they look more like fluctuations, statistically much less significant than that appearing in electron plots

* sure, there is a decrease in the average number of tracks per event with a cut applied.
@JonaJJSJ-crypto can you check if the electron_eta and phi anomaly disappear if you impose a cut of "only events with at least one track with pT>10,15" in the event?

I think this would not affect because as can be seen in the lastly updated study, most of events contain energetic tracks. Almost every event contains at least one track with pt>15GeV.

JonaJJSJ-crypto commented 3 years ago

In the case of second most energetic electrons:

JonaJJSJ-crypto commented 3 years ago

@JonaJJSJ-crypto, when you have this figured out, can you check on the expected size reduction for data and signal files? This should be done for:
* cut on pt for at least one electron

* cut on pt for at least one track

* cut on pt for at least two electron

* cut on pt for at least two tracks
Here I are the expected size reduction implemented all the cuts with pt>10GeV. This cut has been selected because considering that the number of events containing at least one energetic electron this represent a 0.15% reduction on the events in MC, and is a great reduction in Data. In the case of tracks, there is no reduction at all for MC, but it can be seen that this does not affect the reduction as much as the cut in electrons in Data. Is important to notice that for the Doublephoton data set there is no difference between applying the cut to Two electrons with respect to one electron because all the events here contain 2 energetic electrons or photons. On the other hand, there is a difference in the case of tracks but I'm not sure why. In the case of pt>5Gev there actually almost no difference with not filtered AOD.

caredg commented 3 years ago

I have plotted eta vs pt vs chi2, it seems that these anomalies are mostly bunches of tracks grouped in pt and eta, that im assuming are jets. and other that are grouped in chi2 and eta, but im not sure what those are

Could you plot the same graph but only for the two tracks with higher pT in the event (without and with a pT cut of 10, 15 on at least one track)?

caredg commented 3 years ago

@JonaJJSJ-crypto, when you have this figured out, can you check on the expected size reduction for data and signal files? This should be done for:
* cut on pt for at least one electron

* cut on pt for at least one track

* cut on pt for at least two electron

* cut on pt for at least two tracks
Here I are the expected size reduction implemented all the cuts with pt>10GeV. This cut has been selected because considering that the number of events containing at least one energetic electron this represent a 0.15% reduction on the events in MC, and is a great reduction in Data. In the case of tracks, there is no reduction at all for MC, but it can be seen that this does not affect the reduction as much as the cut in electrons in Data. Is important to notice that for the Doublephoton data set there is no difference between applying the cut to Two electrons with respect to one electron because all the events here contain 2 energetic electrons or photons. On the other hand, there is a difference in the case of tracks but I'm not sure why. In the case of pt>5Gev there actually almost no difference with not filtered AOD.

Could you please:

repeat the electron cuts but without the photon cut (no cuts on photons). Forget about the tracks.

JonaJJSJ-crypto commented 3 years ago

I have tested

@JonaJJSJ-crypto, when you have this figured out, can you check on the expected size reduction for data and signal files? This should be done for:
* cut on pt for at least one electron

* cut on pt for at least one track

* cut on pt for at least two electron

* cut on pt for at least two tracks
Here I are the expected size reduction implemented all the cuts with pt>10GeV. This cut has been selected because considering that the number of events containing at least one energetic electron this represent a 0.15% reduction on the events in MC, and is a great reduction in Data. In the case of tracks, there is no reduction at all for MC, but it can be seen that this does not affect the reduction as much as the cut in electrons in Data. Is important to notice that for the Doublephoton data set there is no difference between applying the cut to Two electrons with respect to one electron because all the events here contain 2 energetic electrons or photons. On the other hand, there is a difference in the case of tracks but I'm not sure why. In the case of pt>5Gev there actually almost no difference with not filtered AOD.
Could you please:
* repeat the electron cuts but without the photon cut (no cuts on photons).  Forget about the tracks.

I had to test again this Filter, it seems I have made a mistake when studying of the first time. Good news the cuts in electrons and photons seems to not be relevant at all with respect to the final output.

caredg commented 3 years ago

@JonaJJSJ-crypto, what i want to understand is how (and how often or if) the LW particles are produced against a recoling particle, like jets. Or in another way, how often they are back to back or not.

caredg commented 3 years ago

@JonaJJSJ-crypto, in the same place as always, there is a folder called signalStudy_round5. There there is a merged file called LWSM200DnR_noFilter_notracks.root. This file was obtained without filters and does contain tracks (we do not need them for this test). What I would like to know is, at the generator level, how many events, out of the 150K, we expect to loose if we apply our pre-selection criteria, i.e., we require at least two electrons with pT>15 and at least 4 jets with pT>10. How many we will actually loose (i.e., at reconstructed level?)

caredg commented 3 years ago

@JonaJJSJ-crypto, I would also like to know whether we loose any very displaced event (say with gen vertex position larger than 20microns) after applying the electron or jet pre-selection

caredg commented 2 years ago

There is a new problem with the signal. Looking at the number of primary vertices distribution for the ntuples, we see two bumps.
ntupleNpvProblem

This is also seen in one of the reco files (before producing the nptuples), so it has to do with the simulation. recoNPVProblem

It seems unlikely that this a problem at the generator level, because I do not see any structure in the gen vertex size: genNPVProblem

So everything points out to the minbias mixing again. My suspicion is that the different simulate files for minbias have to different structures. I will make some tests

JonaJJSJ-crypto commented 2 years ago

This

There is a new problem with the signal. Looking at the number of primary vertices distribution for the ntuples, we see two bumps.

This is also seen in one of the reco files (before producing the nptuples), so it has to do with the simulation.

It seems unlikely that this a problem at the generator level, because I do not see any structure in the gen vertex size:

So everything points out to the minbias mixing again. My suspicion is that the different simulate files for minbias have to different structures. I will make some tests

This behavior is also present in the other datasets for different masses but with slightly different shape.Wa there a difference in the sleection of minimun Bias

imagen imagen imagen imagen

JonaJJSJ-crypto commented 2 years ago

@JonaJJSJ-crypto Agregar dsitribuciones de los mejores jets

JonaJJSJ-crypto commented 2 years ago

@caredg Aqui adjuntare las prncipales distribuciones para jets, sin sleecion alguna esta es la distribucion de Jets Pt y Eta para jets reconstruidos y para las hijas de Z.

JonaJJSJ-crypto / Proyecto-de-Tesis

Acausal signal study #35