Mirror of the best candidate showers page.
There was an error in the results from Apr. 11 - I had only changed part of the ArmOfTheCross() code to be optimized for Chiquita. That is fixed here.
Also I realized that on every SOTC try, Rext_Shawn(). I changed this for the moment so that the code first checks all 8 possibilities using Rext_Brant() and then reconstructs the best using Rext_Shawn(). This should eliminate the possibility that re-evaluating the chisqs after calling Rext_Shawn changes the SOTC trial used in the final result.
After making this change, evaluating the chisq values after running Rext_Shawn() updates them in the expected way. The final implementation should probably make a choice based on the relative likelihood of the 8 SOTC trials.
It was noticed in the meeting on Friday that the chisq output is not the same between v3.01 and Dummy reconstructions. It seems that something odd is going on in Rext_Shawn... when the function which evaluates the location chisq is called, it changes the fit results.
The two functions which update the chisq values are fcnChiSqAngle and fcnChiSqCore. The two variables which hold the angle and location chisq values are minchisq and minminusloglik, respectively. I haven't figured out how calling these functions can be changing the fit parameters after the minimization routine has finished.
The following reconstructions show the results of uncommenting either one function or the other in Rext_Shawn (or switching the variable that they save the result to). Version C is the way v3.01 was implemented originally. The DummyReconstruct routine calls a function called UpdateChiSqs, which calls both fcnChiSqAngle(minchisq) and fcnChiSqCore(minminusloglik). Inserting this function directly into Rext_Shawn gives the same results as version E below.
A. Shower 1, version 2.35 Xangle = 0.12, Xloc = 0.68
B. Version 3.01, no chisq evaluation
Xangle = 0.12, Xloc = 0.68
The fit is different but similar.
As expected, the chisq values are not updated from the 2.35 starting point.
C. Version 3.01, fcnChiSqAngle(minchisq)
Xangle = 0.08, Xloc = 0.68
Only the angle chisq is updated (as expected); fit is the same as B.
D. Version 3.01, fcnChiSqCore(minminusloglik)
Xangle = 0.12, Xloc = 30763.36
Only the location chisq is updated (as expected); fit is the same as B.
E. Version 3.01, both of the above
Xangle = 0.08, Xloc = 30763.36
Both chisqs are updated
F. Version 3.01, fcnChiSqAngle(minminusloglik)
Xangle = 0.12, Xloc = 0.08
New Xangle is saved to Xlocation.
G. Version 3.01, fcnChiSqCore(minchisq)
Xangle = 30763.36, Xloc = 0.68
New Xlocation is saved to Xangle.
