Sylvester to Andrea Signori

This apparent discrepancy is caused by the time stability systematic uncertainty, which is (relatively speaking) a bit smaller for K- compared to the other particles. For the other major contributions to the systematic the relative size of the K- systematics is in line with the other particles.

To understand why the K- time stability systematic is smaller, you have to understand how it is estimated (and why it has to be introduced in the first place). The proton data is a combination of data taken in 2000 on unpolarized H, in 2004 with transversely polarized H, and in 2005 with transversely polarized H. The deuteron data is a combination of data taken in 2000 on longitudinally polarized D, in 2004 with unpolarized D and in 2005 with unpolarized D.

Each of these samples was analyzed separately, because the different target setup requires a different MC setup for the unfolding. In the end, the final result is obtained by taking a weighted average of the three samples. Of course, in the course of 5 years some degree of detector instability is to be expected, and it is also to be expected that these fluctuations are not 100% compensated for by the MC.

To estimate the magnitude of the fluctuations, a t-test was performed on the different samples. The results of these test necessitated the introduction of a systematic uncertainty for both pions, and the K+. The K- samples however, where in perfect statistical agreement. Of course, the explanation for this is the much larger size of the _relative_ statistical uncertainty of the K- multiplicities, due to the smaller K- cross section. This is especially true at very low z (below the recommended lower bound of 0.2) and high x (due to the falling exponential shape of the cross section).

The systematic uncertainty is estimated by a multidimensional polynomial fit to the year-to-year discrepancy to avoid statistical fluctuations in the systematic uncertainty. For consistency, we decided to introduce a similar uncertainty for the K-. For the same reason as that the t-test was in agreement, the size of this contribution to the systematic is generally a bit smaller than for the other particles - I say generally because this is still localized to only a couple of kinematic areas.

While this effect is a bit more pronounced for the deuteron results (the shape of the time dependent fluctuations is different between the targets), this is also present for the proton results.

Finally, when looking at the integrated multiplicities - where the K- statistics/bin are substantially higher - you can see that the systematic actually falls in line nicely with the pions and K+. All systematics are _re-evaluated_ for the integrated results (to account for correlations between the systematic uncertainties), and this includes a new fit for the time stability systematic.