Dear ROOT wizards,
I am having problems with TSpectrum and TSpectrum2. About 1% of the time TSpectrum is not finding visible peaks. Also, TSpectrum2 doesn’t ever seem to work. I am using it in compiled code. I attach a snippet of the call:
TSpectrum* s = new TSpectrum(); s->Search(_histoRadius,2.," ",0.5); TSpectrum2* s2 = new TSpectrum2(); s2->Search(_histoRadiusDCA);
please look in the checkhits directory and take a look at event 39. I see an obvious peak. Also, I have run 1000 events, found 9 bad ones total, and don’t see a clear pattern to tell me what characterizes these events.
In the TSpectrum2 calls, look at (in checkhits) at any of the RadiusDCA 2D histos: e.g., _histoRadiusDCA00002. It finds lots of peaks, none of them correct. The histos are created (using tfs) like this:
TH2D* _histoRadiusDCA = tfs->make<TH2D>(radiusDCAHisto.name(),radiusDCAHisto.title(), 80,100.,500.,100,0.,100.);
I’ve tried TH2F and get the same results. One can see obvious peaks in the histogram. I did all this on one machine (ilcsim2.fnal.gov) and then decided to check another machine, so I also ran peak2.C on the original machine and fnalu (flxi05.fnal.gov for specificity) and the results look somewhat sensible, but there seem to be lots of ghosts in the upper right corner of the plots in all cases on both machines. I would think the ghosts should be randomly distributed.
Can someone help me out? Perhaps I need to tweak an argument or I’m doing something obviously wrong. Phillipe Canal is a couple of floors down so that might be a way to get quick info out of me.
–thanks, Robert Bernstein
houghtest_03.root (198 KB)