This study on statistical errors was prompted by the horrible linear fit to the fill-by-fill inclusive electron asymmetry, as shown at the June 2008 Collaboration Meeting at UC Davis.

First, the formula for statistical errors for the spin asymmetry (both single and double) are derived from the asymmetry formulas. (See attachment: Report-July-28-2008.pdf) These derived errors are larger than the errors calculated from the approximate formula, and yield a slightly better fit. However, it's still a very bad fit.

So a couple of things were tried:

• Cut on number of events, to remove short runs
• Updated run list
• Cut on relative luminosity R (i.e. exclude runs/fills with |R-1| > 0.2)

The asymmetry following those cuts are shown below:

By run: A_LL A_B A_Y

The fit is still bad, especially for single spin asymmetries.

Then additional cuts were applied, |Vertex-Z| < 35 and |Track-eta| < 0.7 and a tighter dE/dx cut. The results are shown below:

By run: A_LL A_B A_Y

The fit is significantly improved (especially for single spin), however, the statistics were reduced by a lot. So it's a bit unclear whether the improvement came from the data or simply due to the now much larger error bars.

Since the reasoning behind the tight cuts on Vertex-Z and Track-eta were not entirely clear, we decided it would be best to study those cuts first before going back to the asymmetry. The brief study on those cuts is discussed here: drupal.star.bnl.gov/STAR/blog-entry/kurnadi/2008/aug/11/cuts-primary-vertex-z-track-eta

Latest asymmetry results following Vertex-Z and Point-η cuts: http://drupal.star.bnl.gov/STAR/blog-entry/kurnadi/2008/aug/20/spin-asymmetry-statistical-errors-part-2