Try MET uncertainties due to individual JEC variations

[andrey-popov]

Jet pt threshold was lowered to 15 GeV in commit 4b78a39. This allows propagating JEC into MET without rerunning in Grid, which is needed to change JEC starting from PEC tuples or evaluate individual JEC uncertainties. The drawback is that this has increased the number of stored jets by as much as a factor of two.

Instead, could try evaluating MET uncertainties due to all individual JEC variations in Grid and saving them in PEC tuples. Then the jet pt threshold can be raised back to 20 GeV.

[andrey-popov]

There are more than 30 individual sources of JEC uncertainties. Saving their impact on MET would mean storing more than 120 floats per event — about 500 bytes. At the moment one jet takes 87 bytes, and some fields might be easy to compress. Would have to add more than five jets due to the loosened threshold in order to justify storing all uncertainties in MET.