Daily Notes at LPC
Here you can find a record of almost all my activities in a chronological order since February 25th. Before that I had all my stuff in papers, in post its, in little notes and my mind so I will try to put that information directly in the
CalorimeterTriggerUpgrades twiki.
Monday, February 25th, 2013
First I have to make the necessary corrections in order to get the tree for the zerobias RAW data. Think about these topics:
- If the makeRatesTreesCrab_cfg.py file is changed, the complete CMSSW_X_X_X has to be re built. Sometimes I got troubles when I forget this.
- Is not OK send jobs to the grid if you are not sure about what are you doing or if not sure about your code does
- Test your code with a few data with cmsRun and look if the results is close to what you want before sent your job to the grid
I have to complete the script for do the matching between the reco and the raw files. Choose the best match in the whole loop, keep it in an array, organize the array from minor to major and then choose the smallest value. That is the best candidate.
Have to study the
UCT2015EClusterProducer.cc for tomorrow’s (Tuesday, February 26th) meeting.
Tuesday, February 26th, 2013
Continue the study of the
UCT2015EClusterProducer.cc code for today’s meeting. The main change is in fact a geometry change in the ECAL, because now we want more detail and the seed used is 1x1 instead the 2x1 used before.
The terminal is still joining all the root files created yesterday in the Job 20 (zerobias1234_raw_run198609_tree.root).
The merge process has finished and now I have zerobias1234_raw_run198609 tree.root file in the 20th job in the output folder.
Wednesday, February 27th, 2013
The test with the root file generated yesterday are OK and I guess are ready to put as the input of the matching script ( Status: still under construction!! ).
The main problem is find the correct candidate because it takes a lots of time iterate one by one and choose a match ( Need more ideas )
For today is good if I can develop a script that plots L1, UCT2015 and
STAGE1B in the same plot.
Thursday, February 28th, 2013
The authors of the main project updated to a new version, so I deleted the CMSSW535 and installed again with the new tags ( version V00-02-00 ) and plottersC from Andres’ repository.
At least compile!
Now try to produce some Efficiency plots. I got the skim for makeEfficiencyTree from: /store/data/Run2012C/DoubleElectron/RAW-RECO/Zelectron-PrompSkim-
v2/000/198/230/0000/....root
Until now is a “problem” in stage1B because always is zero the values of Pt.
On the other hand the data for makeRates has to be a zerobias raw because it’s high pile up, so I will try to run that with the data from: /pnfs/cms/WAX/11/store/data/Run2012C/ZeroBias1/RAW/v1/000/198/609/ and try this time with the first file here.
Wrong! There is an error like:
"Running path p1
calling event method for module
RateTree/ ’rlxTauUCTRatesStage1B’
Exception message: cant find the key puLevel in the UCTCandidate, sorry"
Until I can fix that, follow with the Efficiency studies and try to figure it out why that zero appears?
Friday, March 01, 2013
Try to plot the relative efficiencies in the same canvas: Current vs L1 and current vs
L1B
Monday, March 04, 2013
Plots proposed on Friday afternoon are ready, just have to put a different color in each one because they are almost equal. Cannot find the right color command.
The plot contains the efficiency resolutions for the Current stage, for the UCT2015 Stage and for the UCT2015 Stage 1B.
Ready resolutions for Eta, Phi and Pt, but they are very similar, so it is important to set different colors for the superimposed plots.
Command color found but are still so difficult to distinguish each one.
Think about:
- Where -1 comes from?
- Need more statistics!
Tuesday, March 05, 2013
Start again because a problem with the PC.
Now I am using more data in order to get more entries in the histograms and a better statistic.
Now I have two kind of plots for Pt, Eta and Phi:
- In a column, the current on top, Stage1 in the middle and Stage 1B at the bottom.
- A superimposed plot of Stage1 and Stage1B, so you can see the differences.
Improve the script.
Wednesday, March 06, 2013
Now I have all the stuff created yesterday in a script, so, it is easier make test, add and remove lines.
Have to make the normalization of the plots and then compare and evaluate the implications.
Having troubles with the normalization process.
Starting the troubles with my laptop
...
Thursday, March 07, 2013
After a lot a troubles the code compiles and is "almost" efficient. The normalization process is complete and the first plots are ready.
The first three are for a comparative view of the Current stage vs Stage1 vs
Stage1B. The last three are for Stage1 and
Stage1B superimposed in the same plot.
Friday, March 08, 2013
Try some different cuts and axis scales in order to understand the generated plots.
Study the code for plot Efficiencies. Two options, the code provided by the UCT2015 package and wrote in python or the modifications made by Andres in C++.
Monday, March 11, 2013
Make a presentation with the most representative plots obtained last week.
Study
UCT2015Stage1BProducer and
UCT2015ClusterProducer code for tomorrow's meeting
Tuesday, March 12, 2013
Presentation completed. Received comments, changes and topic to improve the current study.
How about the new seed in
Stage1B ? How many towers has?
Make more test with the "same treshold" for
UCT2015Producer and for
UCT2015Stage1BProducer but there is something that do not work.
Using edmConfigEditor ( plus a py configuration file ) for make the changes and try to find out the error.
Wednesday, 13 March, 2013
Finally arrived my new laptop!!!
Using the day to install Linux, all the necessary tools and try to recover the data from the old one ( RIP )
Thursday, March 14, 2013
Some time at library reading about beta decay, pi mesins, muons, taus and detector's physics. Interesting topics. More lecture is necessary!
Continue the test changing some parameters on configuration files for a electrons skim.
Not good results until now...
Finally a good run!
Friday, March 15, 2013
These plots are the result of run Stage1 and
Stage1B in the "same configuration".
EtaResolutionL1L1B-SI4.png
PhiResolutionL1L1B-SI4.png
PtResolutionL1L1B-SI4.png
As we can see, in Pt resolution the shift between the two stages is bigger than before. This results were the result of change some tresholds. This changes has the purpouse of give the same parameters for the Stage1 and for
Stage1B.
More study is needed to determine the origin of that shift...
In other hand I prepared the same analysis with isolated branches ( every analysis until now was realized with relaxed branches ) and the resultas present the same shift. The outputs before make the changes for have the "same tresholds" are:
IsoEtaResolutionL1L1B-SI.png
IsoPhiResolutionL1L1B-SI.png
IsoPtResolutionL1L1B-SI.png
and the output after make those changes are:
IsoEtaResolutionL1L1B-SI1.png
IsoPhiResolutionL1L1B-SI1.png
IsoPtResolutionL1L1B-SI1.png
and as the same as we saw in the relaxed case, the shift between Stage1 and
Stage1B increases.
Monday, March 18, 2013
In order to study the isolation variables I have to run the
Stage1B Producer in grid using a skim of electrons ( zelectrons_rawreco__files_2012C-V2.txt ) and a background sample ( minimumBias_2012D_HL_RawReco.txt ).
I had to modify the code for have four more histograms as an output of
UCTStage1BProducer. The isolation variables are:
- relativeJetEMIsolation
- relativeRgnEMIsolation
- relativeJetIsolation
- relativeRgnIsolation
Now with both outputs I can create an script for analyze the isolation variables for the electrons and for the background at the same time, compare and plot them together.
The first plots:
- egRelativeEMJetIsolation.png
- egRelativeEMRgnIsolation.png
- egRelativeJetIsolation.png
- egRelativeRgnIsolation.png
After some plots and weird results I realized that those variables were not enough, so was neccesary add the variables that compute that ones are:
- emCluster
- jetEMIsolation
- rgnEMIsolation
- jetIsolation
- rgnIsolation
Additional plots:
- emCluster.png
- jetEMIsolation.png
- rgnEMIsolation.png
- jetIsolation.png
- rgnIsolation.png
Final plots:
- IsoVar-EGRelativeEMJetIsolation.png
- IsoVar-EGRelativeEMRgnIsolation.png
- IsoVar-EGRelativeJetIsolation.png
- IsoVar-EGRelativeRgnIsolation.png
After this plots I can confirm that something is wrong! Shows negative quantities, but all has to be possitive ( at least in this case and omitting the default/debugging values ).
Go back to the code, search those variables, follow the track and try to understand the results. Pending more analysis...
Starting the analysis between Current Stage, Stage1 and
Stage1B.
The result was not what I hope. Definitely there is something wrong with the code.
- Efficiency-ForCurStg1Stg1B.png
Tuesday, March 19, 2013
Start looking the bugs in the code.
Meeting with Andres, Fan, Kalanand and Sergo for present the results and try to find out why the plots are like this.
Some ideas about how to find the errors:
- wrong comparisons
- logical mistakes
- "obvious" bugs
- wrong thresholds
Study the code and prepare a list of questions for the meeting with Sridhara Dasu tomorrow.
Wednesday, March 20, 2013
Meeting with the UCT2015 developer Sridhara Dasu, Andres, Kalanand and Sergo. Present the results to him and try to understand what is happening?
The result of the meetong is that in fact the code was wrong and has to be corrected.
Prepare and send an email to the person who works with the CVS.
Thursday, March 21,2013
Some parts of the code was fixed and now I have to repeat the studies, runs and get the same plots.
First start in a new area and install all the necessary code ( CMSSW_5_3_5 ).
After running my scripts, at least in efficiencies, the plots looks fine:
- Efficiency-CurStg1Stg1B-01.png
In the isolation variables still is something that do not work well:
- IsoVar-EGRelativeEMJetIsolation-01.png
- IsoVar-EGRelativeEMRgnIsolation-01.png
- IsoVar-EGRelativeJetIsolation-01.png
- IsoVar-EGRelativeRgnIsolation-01.png
Friday, March 22, 2013
Study of isolation variables and try to understand:
- why we get negative values?
- why apears a shift?
Monday, March 25, 2013
Look into the size of each isolation variable and the way it is defined:
- rgnEMIsolation: 4 x 4 / 3 x 3
- jetEMIsolation: 12 x 12 / 3 x 3
- rgnIsolation: 4 x 4 / 3 x 3
- jetIsolation: 12 x 12 / 3 x 3
Tuesday, March 26, 2013
The code was modified again and looks great.
Finish the efficiencies plots and starting looking the rates.
Wednesday, March 27, 2013
Looking for information about Hadron Outer scintillators.
Prepare a *.tar file with all the plots I got on the studies and send to Andres for make a pdf and send to Sridhara and Evan ( responsives for the code ) for comments.
It looks really well!
Thursday, March 28, 2013
Plot many comparatives in isolation variables with a skim of electrons and modify the script until get the correct plots.
To get this, I had to send to the local grid. Waiting for the results.
Friday, March 29, 2013
Work with the results got yesterday and now make a similar analysis in the grid, but this time for a minimum bias sample.
Waiting for the results.
The new plots for isolation variables are:
- IsoVar-EGRelativeEMJetIsolation-02.png
- IsoVar-EGRelativeEMRgnIsolation-02.png
- IsoVar-EGRelativeJetIsolation-02.png
- IsoVar-EGRelativeRgnIsolation-02.png
now we see that all of them are completely possitive.
Monday, April 01, 2013 - Friday, April 05, 2013
This week is for repeat the performance studies in a final state, i.e, with all the UCT2015 Code in the last version and without buggs ( I hope so! )
This test includes plots for Efficiencies, Rates, Resolutions ( of Pt, Eta and Phi), Isolation variables and a final presentation to the L1 Trigger Upgrade Studies Group at
LPC on Monday 8.
The main difference between this last study and the older ones is the size of the processed data. In the previous versions we deal with relative small samples of particles ( skim of electrons, taus, minimum bias, zero bias ) and now we have big files.
The scripts has been corrected and runs really good.
Monday, April 08, 2013
A close look to the basics of HO Studies.
I found a paper ( from Fermilab ) that gives a brief description about the Hadron Outer scintillators but need more information.
Another necessary topic is the Particle Gun and here I can find a first overview (
https://twiki.cern.ch/twiki/bin/view/CMSPublic/SWGuideParticleGuns )
Tuesday, April 09, 2013
Presentation about performance studies to the L1 Trigger Upgrade Studies Group.
The conclusion is that could be interesting if we can get our own WP ( working point ) making some changes in the default values in Isolation Variables. The idea is make plots for efficiency and rates for each change that we do, then we can compare the different plots and chosse the ideal value for the variables we are interested in.
Wednesday, April 10, 2013
The efficiency plots are ready and I can choose my "favorite candidates" until now but I have to wait for the rates plots.
I have some troubles with the rates. The script for rates works really good for isolated and relaxed branches and, theoretically if I use the relax branches and then apply some isolation cuts I have to obtain a similar result as if I work with isolation branches directly, but I am not getting that!
Thursday, April 11, 2013
I am still working on the rates script and triying to figure it out why the plot fails. Maybe the tree does not contain or is not keeping the correct information.
Friday, April 12, 2013
Finally the script for the rates is complete!
To access to "isolation variables" is necessary make some modifications in
UCTRateTree.C file. First you have to add additional variables for represent what you want to access and then treat them as the other variables you find there.
Monday, April 15, 2013
Prepare a presentation based on the different plot for Efficiencies and Rates in order to choose the best working point.
Make a table with the values of Efficiency and Rates for 20 and 30
GeV
- EG-rlx-Eff-IsoVar-20GeV-001.png
- EG-rlx-Eff-IsoVar-30GeV-001.png
- EG-rlx-Rat-IsoVar-20GeV30GeV-001.png
Tuesday, April 16, 2013
Feedback. We almost have a good working point and until now we have some candidates ( 2 or 3 ).
It is necessary to add the
PileUp Subtraction criteria for improve the working point candidates.
Wednesday, April 17, 2013
Add the
PileUp Subtraction to the rates script.
The is some problem but it is hard to find because the code compiles well and there is no error or warning.
Thursday, April 18, 2013
Stil having troubles with the rates, they do not produce what I want.
Finally the rates script is ready!
Now, add the
PileUp subtraction to the Efficiencies script.
Friday, April 19, 2013
Prepare a presentation for the
LPC L1Trigger Upgrade Studies Group (Andres, Fan, Kalanand, Sergo)
Presentation is ready for today's meeting.
- EG-rlx-Eff-IsoVarPUS-20GeV-001.png
- EG-rlx-Eff-IsoVarPUS-30GeV-001.png
- EG-rlx-Rat-IsoVarPUS-20GeV30GeV-001.png
Monday, April 22, 2013 - Friday, April 26, 2013
Modifications, improvements, details, labels, axis names were made to the presentations.
Monday, April 29, 2013 - Friday, May 03, 2013
This week was dedicated most of the time to learn how to develop a "producer" in order to make some progress in the HO Project.
For the
L1Trigger Upgrade Project this week was really important because I made a lot of progress. I made a lot of debuging in my own code and also I divided it into small scripts and that makes more modular the production of plots for performance studies.
Finally on friday I made a presentation for the
LPC L1 Trigger Upgrade Studies Group and as a result I have more work to do looking at iso branches and maybe taus.
Also on friday we said good bye to a good man, teacher and friend, Andres Osorio... Thank you very much for all the patience, concern, advices, friendship and help...
I want the best for you and your family. Good luck in your new project!
Monday, May 06, 2013
Looking for a data sample that contains:
There are some runs that may contain high pileup ( not sure ): 179828, 198588, 198603 and 198609.
Tuesday, May 07, 2013
Now looking for a data sample that contains:
- Zerobias
- RAW-RECO
- High Pileup
After many attempts looking for these characteristics I got a ".root" file but it is less than useless because the file has not enough entries. So, now the objective is find more data.
Wednesday, May 08, 2013
Is it possible combine plots from different datasets? Look some references.
Thursday, May 09, 2013
In order to understand the behavior of the variables that we are studying will be useful include a high pile up sample in the current plots. The problem is that we do not have RAW-RECO data with high pile up. The solution is make a combination of plots from different datasets in the same canvas.
Friday, May 10, 2013
Read the new documentation for the HO Project
Monday, May 13, 2013
Start reading "Measurement of inclusive W and Z boson cross section in pp collisions at sqrt(s) = 8
TeV". The paper version, the PAS version and the analysis notes.
Tuesday, May 14, 2013
Reading day: "HO material" (six papers until now) and "cross section analysis" (three documents) material.
Wednesday, May 15, 2013
Prepare a list of questions about HO and cross section analysis.
Thursday, May 16, 2013
Meeting with the
LPC L1 Trigger Upgrade Studies Group. Discussion about new modifications in the code for have "better" working point candidates.
The proposed changes are:
- relax the pileup cut (5GeV instead of 10GeV)
- flip the efficiency plots
- make the table of results just for 20GeV
- the new values for the isolation variables are: 0.05, 0.1, 0.15
Friday, May 17, 2013
HO reading and some cross section analysis.
Monday, May 20, 2013
Start the production of the new trees for relax the pileup subtraction. Prepare the input files, the configuration files, the submit script and send all the stuff to run in the grid.
Tuesday, May 21, 2013
Continue with the production of the trees.
Monday, July 01, 2013
In the pasts weeks, after some presentations to the L1trigger Upgrade Studies Group we got some conclutions:
- the cuts applied to the I_J and I_R variables were good and help us to discard I_R, because its behavior does not affect (affect in fact, but in a really tiny scale) the performance of the trigger, so now, all the efforts are focused in I_J
- something is weird in the rates plots. The plots with the cuts in I_J are consistent and shows the behavior that we are expecting, but the rates plots are "over/mis" measured and shows numbers that to high/low, depending of the case, so the plan is to surf trough the code and figure it out if in some place we are including some extra cuts or maybe to in the wrong way the integration
- also plot some ROC curves to show the behavior of the efficiency of the signal versus the efficiency of the background. To compute the efficiencies is necessary to use a skim of electrons (signal) and a minimumbias sample (background)
The way to compute the rates is the following (this is a brief description of what does the algorithms):
1. Iterate over events
2. Find the level-1 objects. Here we have three posibilities:
- do not exist level-1 objects
- exist only one level-1 object
- exist more than one level-1 object
and of course, we have three possible responses:
- skip the event
- pick this one
- pick the one with the highest Pt
3. Check the Pt of the level-1 object: Pt (electron or gamma) > 20 gev && I_J < 0.1
4. Plot Pt (just for some help at this point)
5. Integrate the Pt of the level-1 object
6. Fill an histogram with the integration values.
Tuesday, July 02, 2013
In order to finish the roc curve task I need efficiency trees with more events, because there is not enough statistics to see any behavior. The way to do that is generate those trees with a new dataset for electrons and minimimbias. So the plan is find some good root files that contains the information that I need.