[Halld-offline] NAN values for GetPull();

[Blake Leverington]

Hi guys,

Figured out what my problem was.  Since I am putting in the beam and 
target with zero error the pull blows up for these parts of the error 
matrix. So it was just operator error on my part. I hadn't realized 
there were so many pull values. The pull values for the photons in my 
fit were not zero.  (They have a sigma of ~2, not 1, but this is before 
any cuts on probability.)  I think I learned more about the kinematic 
fitter trying to hunt this down than I knew previous to this so all was 
not wasted.

Thanks for the clues about where I should go looking. In the old 
examples given that follow the CMU fitter, the covariance matrix is put 
in by hand or from file, but in GlueX, the kinematic data for each 
particle contains the error matrix for itself and the covariance matrix 
is built from that in the kinematic fitter.

The feedback was appreciated! Thanks,
-Blake

On 09/03/10 10:08 AM, Craig Bookwalter wrote:
> Hi Blake,
>    I'm using the CMU kinematic fitter for some CLAS data, and I've had 
> a similar problem to what you're having. It turned out my input 
> covariance matrix contained some outlandish values (like 1e308's and 
> similar) and it was causing the fitter to spout nonsense. I'd suggest 
> looking at the values of your covariance matrix and input vectors and 
> making sure they're not corrupted somehow before they make it into the 
> fitter. My $0.02.
>
>    --cb
>
> Curtis A. Meyer wrote:
>> Hi Blake -
>>
>>    we started playing with this again at CMU, but unfortunately the code
>> has been orphaned since Matt Bellis left a couple of years ago. I am 
>> a bit
>> surprised that it stopped working as I do  not believe that anything 
>> in the KINFIT
>> code has changed.
>>
>>    curtis
>>
>> On 3/8/10 5:24 PM, Blake Leverington wrote:
>>> Hey all,
>>>
>>> Any ideas on why kfit->GetPull(0); in DKinFit returns NAN though
>>> kfit->Prob(); still returns reasonable values? I remember this working
>>> months ago, put I can't get the pulls from the kinematic fitter 
>>> anymore,
>>> and I think I need to see this for my analysis at the moment.
>>>
>>> Who is maintaining the kinematic fitter now?
>>>
>>> -Blake
>>> _______________________________________________
>>> Halld-offline mailing list
>>> Halld-offline at jlab.org
>>> https://mailman.jlab.org/mailman/listinfo/halld-offline
>>>
>>
>>
>
>