[G12] Kinefitter

[Pedersen, Christopher]

I should clarify that I am not saying that Dustin has claimed that this is tuned for __multi = false For g12. The similarities between the experiments may allow for a satisfactory tune by simply scaling the parameters from another and as long as the pulls and confidence level look good for a clean channel and do not vary from sector to sector or over theta then the fitter should be working correctly as these are what the parameters are binned in. Getting every parameter again from data as MK has insisted may be the most thorough way, but may not be necessary if everything looks fine and does not vary in these variables. Also the scaling of the p_offset parameter or really many parameters in the inside the program should not matter if the fitter is entirely retuned because this change will simply be reflected in the parms files, but because the included parameters for protons and pions were not tuned like this it threw things way off.

Chris
________________________________________
From: g12-bounces at jlab.org [g12-bounces at jlab.org] on behalf of Pedersen, Christopher [cep11k at my.fsu.edu]
Sent: Friday, May 23, 2014 12:59 PM
To: Michael C. Kunkel; g12; Volker Crede
Subject: Re: [G12] Kinefitter

Yes what you sent me is very similar to what is on Dustin's web site except for additions that handle electrons and that in the __multi = false case for some reason p_offset is scaled by 2.5 in the code. This is one of the factors I had to adjust that you claimed you had never heard of someone changing before. Here is the only output from diff between what you sent me and what is on Dustin's web site that doesn't involve additions for electrons:

282,283c285,286
<        p_offset = pPars[part_index][sector-1][bin][2];
<        V(3*i+1,3*i+1) += p_offset*p_offset + 2*p_offset*sqrt(V(3*i+1,3*i+1));
---
>        p_offset = 2.5*(pPars[part_index][sector-1][bin][2]);
>        V(3*i+1,3*i+1) += p_offset*p_offset + 2*p_offset*sqrt(fabs(V(3*i+1,3*i+1)));

Where the top is directly from Dustin's web site and the bottom is what you sent me. This is almost certainly the reason that the momentum pulls using the fitter as you sent it are too narrow and the rise in the confidence level toward 1 as this significantly increases the entries for momentum in the covariance matrix.

Chris
________________________________________
From: g12-bounces at jlab.org [g12-bounces at jlab.org] on behalf of Michael C. Kunkel [mkunkel at jlab.org]
Sent: Thursday, May 22, 2014 11:52 PM
To: g12; Volker Crede
Subject: [G12] Kinefitter

Greetings,

I wanted to point out that the fitter FSU has and g12 has been using for
multi = true has been tuned by Dustin himself.

Please read
https://userweb.jlab.org/~dustin/software/Makeshift/HowTo/G12-Testing/

I know I said Georgie tuned it, he did, and found no discrepancy with
Dustin's tune.

BR
MK
_______________________________________________
G12 mailing list
G12 at jlab.org
https://mailman.jlab.org/mailman/listinfo/g12
_______________________________________________
G12 mailing list
G12 at jlab.org
https://mailman.jlab.org/mailman/listinfo/g12