Finding hit cluster

[geniejhang]

Recently I wrote hit clustering task and the method is following.

1. Cut the active volume into thin space along to drift direction(y). That is, the active volume is divided into many *xzdy** space. All the hit clustering method is applied in each sliced thin space.
2. Find a pad having maximum hit charge. Let this hit be A.
3. Find hits on neighboring 8 pads. Let there hits be B.
4. If the hit A is having the largest charge than B, calculate mean of the hits A and B weighted by their charge.
5. If there's the other hit having charge larger than A, let the hit be A, go back to 3 and start over.
6. Do this until all the hits are clustered.

I think this method is not that bad, but it doesn't take into account of the position errors of the hit. So, this method apparently needs to be enhanced. Please give any comment or suggestion.

I'm trying to look up this book, Applied Multivariate Statistical Analysis. Maybe later.

[geniejhang]

Here's one extreme example of current hit clustering code. Red dots are hits and blue spheres are cluster point.

The above image shows an event most of the pads fired and those hits are clustered.

Below image shows only two hits are clustered.

Definitely we have to improve the clustering algorithm. In order to do that, we need some MC data.

[geniejhang]

Star cluster finder and hit finder note.

• https://drupal.star.bnl.gov/STAR/files/starnotes/sn0238.ps.gz