What is the significance of a slow turn-on curve in trigger efficiency plots?

[ChrisVer]

I am trying to understand how a Trigger efficiency plot should be interpreted in the scenario of an analysis that is using that trigger.
What does a turn-on curve tell us about events that pass the trigger?

One example (I think it's PUBLIC so it's accessible): https://twiki.cern.ch/twiki/bin/view/AtlasPublic/MissingEtTriggerPublicResults
or in particular a plot like this: https://twiki.cern.ch/twiki/pub/AtlasPublic/MissingEtTriggerPublicResults/Preliminary3_Wmunu_L1.png

In the plot one can see that the HLT_xe70 (that works together with the L1_xe50) efficiency is not that well where it's supposed to be (@70GeV). That means that a lot of events that don't pass the trigger requirements, actually do... is that the right way to read it?
But also what is that curve telling me of an event with Etmiss of let's say 90GeV and an event of 30GeV that actually fire the trigger?

 

[mfb]

A lot of events that have e. g. 70 GeV missing transverse energy in the offline reconstruction don't pass the trigger requirements, right. I'm not sure how to interpret the "no muons" for the x axis.

But also what is that curve telling me of an event with Etmiss of let's say 90GeV and an event of 30GeV that actually fire the trigger?

Where does that 30 GeV value belong to?

 

[ChrisVer]

they label the "without muon corrections" as no-muons I guess... now what is the muon corrections- that I am not sure about...

Where does that 30 GeV value belong to?

if you have some data and apply the trigger you may at the end have an event with MET=30GeV that happened to pass the trigger requirement. This confuses me. In fact the slow turn-on curve gives me the impression that a lot of trash can be there below MET ~160GeV.
As a result I don't understand why would someone use that trigger if it's going to provide him with such bad events in such a wide range, and not instead ask 'offline' a MET>70GeV.

 

[mfb]

if you have some data and apply the trigger you may at the end have an event with MET=30GeV that happened to pass the trigger requirement.

It is very unlikely, as the curve at 30 GeV suggests.

As a result I don't understand why would someone use that trigger if it's going to provide him with such bad events in such a wide range, and not instead ask 'offline' a MET>70GeV.

The better offline reconstruction takes much more time than the trigger has. The trigger is much faster and less precise, and missing Et is one of the variables where this hurts most. The analyses usually take offline cuts significantly above the trigger threshold for that reason.

 

[ChrisVer]

It is very unlikely, as the curve at 30 GeV suggests.

well unlikely but also there is a very large number of events with low missing energies...

The trigger is much faster and less precise, and missing Et is one of the variables where this hurts most. The analyses usually take offline cuts significantly above the trigger threshold for that reason.

well that is true...but, I had the problem with trying to estimate the multijet background using a low Etmiss (<100 GeV) control region when also applying the trigger... some people complained about it (due to this slow turn on curve) and that's why I am trying to understand better what those plots mean and what exactly are those let's say "trash" events which fired the trigger without satisfying the 70GeV requirement, or even did but were still on the turn-on region...

 

[mfb]

I guess you need some missing Et expert for that. One important question would be how muons are treated. Apart from the neutrinos they are the only particles without (relevant) energy deposition in the calorimeters. In leptonic W decays as pictured here, muon and neutrino tend to have opposite momenta. If the trigger misses the muon, its reconstructed missing Et is much smaller than the offline value.

A precise estimate of the jet energy is probably also a challenge in the trigger.

 

[ofirg]

well that is true...but, I had the problem with trying to estimate the multijet background using a low Etmiss (<100 GeV) control region when also applying the trigger... some people complained about it (due to this slow turn on curve) and that's why I am trying to understand better what those plots mean and what exactly are those let's say "trash" events which fired the trigger without satisfying the 70GeV requirement, or even did but were still on the turn-on region...

You can't use a 70GeV trigger EtMiss requirement for a control region with EtMiss<100GeV at the oflline level. Most of your events would not pass the trigger. You have to look for a different trigger for these events in the control region. It can be a prescaled trigger if need be.