Meeting No. 2
Data Analysis:
This week I finished the simulation of the 195MeV data with
500M events. The final results are shown in plot below. Now
I am running simulation for the 330MeV setting. The Simulation
for 495MeV setting will follow at the end.
Before reporting final results I need to investigate again the
systematic uncertainties.
Simulation for Lupolen target
I am still working on the full simulation with the Lupolen target.
I want to see if I can use this target for the precise determination
of the Hydrogen spectrum, although the Carbon/Hydrogen rate ratio is
80%/20%.
Optics for Spectrometer A:
While I was waiting for the analysis to finish, I revisited my analysis
of the optics data taken in 2013. I have already determined the matrix for
spectrometer B. Now I also want to get a new matrix for spectrometer A.
I investigated my code and realized that I have the wrong formula for calculating
the integral of the F-distribution. The number of entries was missing in the
denominator. I have corrected this and now the values are reasonable.
My code works works iteratively, such that in each iteration rejects the
element with largest p-value. This is done so long until all the values
are below a certain threshold (5%). This is similar to dmaini, but dmaini
does not work iteratively, but in one swing throws away all the elements,
that are larger of certain values. I have also wrote such "fast" algorithm
to test the SVD code, but this kind of procedure does not return the best
result, because when removing one element the weights for the remaining
elements also change. The animation below shows the iterative process
of finding the matrix.
The problem with this code is, that if we want to have a good matrix, we
need to work with a very large matrix (~4GB) and it takes a lot of time
to come to the end. In the next step I need to see, if the results
I get are reasonable and determine the final resolutions.
