I describe a small (1 line) change in the current angle fitting routine. This change was motivated by what I thought was the advantage of the likelihood fitter over the current chi-squared fitter: The ability to use the knowledge that light can not come early. To take advantage of this fact I simply keep all early hits - early with respect to the fitted plane.

Below I simply show some plots that demonstrate the improvement in angular reconstruction. One should compare these plots to the plots shown in my previous memo. To ease the comparison I have overlayed the results from the online fitter over the results of the new fitter in all of the relevant histograms. The dashed lines are the current online fitter. If you have difficulty seeing which lines are dashed the poorer results are from the online fitter. The 2 sets of results came from the same run, the same calibrations, and the same DontFits file.

In general I find a 20% improvement in DELEO. This brings the chi-square fitter to within about 10% of the resolution of the likelihood fitter. I would be surprised if further refinements could not improve the resolution by this amount.

Stefan has taken a quick look with the simulation and sees a similar improvement on gamma showers.

Deleo: ALL DATA (90% of triggered events). (Online plot)

Deleo: NFit > 40 (30% of Triggered Events) (Online Plot)

Deleo vs. NFit (Online Plot)

Deleo vs. NHit

Deleo vs. Zenith Angle (All Events) (Online Plot)

Deleo vs. Theta (NFit > 40) (Online Plot)

Delta Angle - Between current online fitter and this fitter.

Conclusions:

The "improved" angle fitter takes advantage of the fact that for Milagro the main source of poor angular resolution is late light. To first order all early light is good. By implementing this knowledge in the angle fitter DELEO improves by about 20% (independent of shower selection). This improvement arises from a reduction of the tail in the resolution function. The peak of DELEO does not shift. this is similar to the result obtained using the maximum likelihood angle fitting technique. The mean DELEO over all showers with the current online fitter is about 5.2. With the new fitter this drops to about 4.1. This is not too much worse than the results from the likelihood fitter (10%).