gtexpcube2 does not use the evtype selection from the counts map file

[ianholst]

I was doing an analysis of the all-sky Fermi data, and when selecting events with gtselect, I specified evclass=1024 for ULTRACLEANVETO and evtype=32 for PSF3. A few steps later, I created an exposure cube with this command:

gtexpcube2 \
    infile=livetime.fits \
    outfile=exposure_ULTRACLEANVETO_PSF3.fits \
    cmap=counts_ULTRACLEANVETO_PSF3.fits \
    irfs=P8R3_ULTRACLEANVETO_V3 \
    bincalc=CENTER

However, I noticed that the flux (counts/exposure) was lower than I expected. To investigate, I re-ran the analysis without selecting PSF3 events (I used evtype=3 to select all evtypes). I found that gtexpcube2 produced the same exact results as before.

This is unexpected because the manual for gtexpcube2 says that it will use the evtype specified in the "input file", which I assume means the binned counts map file:

evtype [integer]
    The evtype to be used in generating the background data. The
    default is INDEF which will use the default in the input
    file. This can be overridden by entering the desired event
    type e.g. 3 for FRONT + BACK events.

However, the following command which manually specifies the evtype produces different output.

gtexpcube2 \
    infile=livetime.fits \
    outfile=exposure_ULTRACLEANVETO_PSF3_evtype32.fits \
    cmap=counts_ULTRACLEANVETO_PSF3.fits \
    irfs=P8R3_ULTRACLEANVETO_V3 \
    bincalc=CENTER \
    evtype=32

As expected, the resulting exposure is around a factor of 4 times smaller than when all evtypes are selected.

It looks like either the documentation is incorrect about the behavior when evtype=INDEF in gtexpcube2, or this is a bug.

[nmirabal]

The LAT data is partitioned into 3 different ways.
1)FRONT/BACK
2)PSF types
3)EDISP types
The evtype regulates the type of partition/choice you are want. Useful tables are in the following link:
https://fermi.gsfc.nasa.gov/ssc/data/analysis/documentation/Cicerone/Cicerone_Data/LAT_DP.html

When you do evtype=3, you are basically choosing all the photons as it adds FRONT=1,BACK=2,FRONT+BACK=3. When you do evtype=32 you are only choosing the fourth (best) quartile, not the whole set so the resulting gtselect file will be smaller. The equivalent to evtype=3 would be choose all PSFs by adding (from the Table) 4+8+16+32 or evtype=60. For EDISP 64+128+256+512=960.

[ianholst]

Thanks, I am aware of that behavior. To clarify, the problem is that the exposure does not change when you gtselect a more restrictive evtype.

[nmirabal]

Yes, you are right. You have caught a bug. We will modify the documentation for now and will start thinking of a fix. I have changed the label. Thanks for reporting this.

[jasercion]

In the short term I've updated the parameter file for gtexpcube2 to eliminate the default INDEF input and added a note that evtype must be specified. 1e1f75a
I'll work on implementing consistency checks to make sure that the selected evtype matches the selection chosen for the input counts map and see why these weren't working previously.

[jasercion]

Short term fix implemented and included in release 2.5.0. Proceeding with a comprehensive fix for the next patch release.