I am adding two TProfile’s using the add() function, but the errors being produced in the addition seem to be amplitude dependent such that at the peak of my distribution, the error is largest.
This is not what I would expect to be a reasonable result of adding errors in quadrature. The errors in both TProfile’s are of the same order of magnitude so I would only expect an increase in the error by a factor of root 2 after summing.
Can someone explain how these errors are summed?
TProfile::Add acts like a merge. The two profiles are merged and the resulting error will be then smaller due to the larger statistics. In a TProfile the error depends on the number of entries in the bin and the spread, has nothing to do with the amplitude.
See also my previous posting at the end of root.cern.ch/phpBB2/viewtopic.php?t=8879
Ah, I see. So it just merges the sum, sum of the squares, and the total entries per bin and recalculates the mean and sigma.
That would explain the amplitude dependence I see. I’m trying to use Add to sum two TProfiles with different shapes. One is flat and one is a pulse.