RE: TProfile means problem

From: Snyder, Arthur E. <snyder_at_slac.stanford.edu>
Date: Sun, 10 Jul 2011 00:23:44 -0700

Never mind! I worked out my mistake with boundaries --

Sorry for the noise .. Arthur S.

---

From: owner-roottalk_at_lxroot01.cern.ch [owner-roottalk_at_lxroot01.cern.ch] On Behalf Of Snyder, Arthur E. [snyder_at_slac.stanford.edu] Sent: Saturday, July 09, 2011 11:40 PM
To: Rene Brun
Cc: roottalk_at_pcroot.cern.ch
Subject: [ROOT] TProfile means problem

Hello Rene et al.,

I'm having with |TProfile|. When I plot <y> vs. "x" with |TProfile| I get values in each x-bin that are ~2 (not exactly 2) times bigger than when I just histogram y and read the mean <y> off the histogram.

Has this happen'd before? I don't see where |TProfile| is getting the number it is plotting. It is not, e.g, the RMS. The ratio to the mean is 2.15 -- not exactly 2.0.

I tried to get the y-distribution |TProfile| uses, but can find a function that extracts it. Perhaps |TProfile| uses big y-bins and calculating means from bin centers rather than event average. Is that possible? This doesn't seem to be what the documentation says though ...

The distribution of y is steeply falling from zero -- roughly exponential. The statistics are large even in each x-bin.

?

-Arthur S.

---

Received on Sun Jul 10 2011 - 09:24:41 CEST