From the joint run this morning:
When the profile gets too wide, the fit seems to mix up the gaussian and the quadratic. The background sometimes becomes negative, among other problems. The sigma can become negative!!! This is insane. So I changed NewProfile to have limits on the area, sigma, and mean of the gaussian. It didn't help the backgrounds any.
I created my own function and tried to use TF1 with my routine as a functor. It seems to create huge mean errors and gigantic sigmas, so I don't know what's going on. I backed off of that fit and NewProfile is using merely the limits now. I will have to create a local testbed using the root file and DLC 70.
Modified 21-July-2009 at 11:32
http://hep.physics.wisc.edu/~jnb/cms/21Jul2009
| Previous notes | Next notes | Main slide directory |