Year Dependence

Aus HERMESwiki
Zur Navigation springen Zur Suche springen

Page maintainer: Luciano

Checked.png This page is considered done. It been reviewed by Markus. There may be missing elements, but they are all flagged and the text has no errors.

Intro

Most of the analyses are based on data recorded over several years of data taking. During such long periods, changes in the HERA storage ring (e.g. lepton beam charge, longitudinal polarisation or beam slope position(s)) or the experimental apparatus (e.g. in the detector efficiencies or in the alignment of the spectrometer) could influence the final result. Therefore, when combining data collected in different running periods, the analyser has to check their compatibility. In the evaluation of systematic discrepancies between the different data samples one should keep in mind that, in general, data productions of different years contain completely different amounts of data.

Changes in the experimental setup

Before doing any studies one has to be aware of changes between certain data sets. Changes can arise from both the beam conditions (e.g. beam charge, beam polarization, beam slope) and the experimental setup (e.g. target gas, target polarization, performance of various detectors, trigger efficiencies, TMC, spectrometer misalignment, etc).


The main periods of data taking are, e.g., reviewed in


A detailed list of the various changes in the experimental setup of the HERMES experiment is available at


The changes to the beam, target and triggers are detailed on their respective pages

Methods

In order to quantify the impact of possible "Year Dependence" effects on our results one should apply a proper statistical test. Statistical tests, like the t-student test and the chi^2 test, allow to quantify the agreement/disagreement between different data sets. In general, two data sets can be considered compatible if the differences are smaller compared to the statistical uncertainty. In this case one can conlude that the observed differences are consistent with the statistical fluctuations and the two data samples can be safely merged. If, on the other hand, the differences between two data samples are systematically larger than the statistical uncertaities, then the two samples are likely incompatible and cannot be merged. In particular, one usually has to fix a confidence level (e.g. 90%, 95%, 99% C.L.), and then express the result of the statistical test in terms of agreement/disagreement within the confidence level. For instance, if the t-student test succeeds (fails) at 95% C.L. one can state that the two samples are statistically compatible (incompatible) with a C.L. of 95%. If the two samples are found to be incompatible one should investigate the origin of the incompatibility by performing dedicated tests/comparisons on all the possible sources (see examples reported in the "Changes in the experimental setup" section above). Please refer to the Statistical tests section in the HERMES WIKI and Statistical hypothesis testing WIKIPEDIA for further details.


A very simple and fast way to check if the final results (e.g. azimuthal moments ) obtained separately from two different data samples Fehler beim Parsen (MathML mit SVG- oder PNG-Rückgriff (empfohlen für moderne Browser und Barrierefreiheitswerkzeuge): Ungültige Antwort („Math extension cannot connect to Restbase.“) von Server „https://wikimedia.org/api/rest_v1/“:): {\displaystyle D1} and Fehler beim Parsen (MathML mit SVG- oder PNG-Rückgriff (empfohlen für moderne Browser und Barrierefreiheitswerkzeuge): Ungültige Antwort („Math extension cannot connect to Restbase.“) von Server „https://wikimedia.org/api/rest_v1/“:): {\displaystyle D2} are compatible, consists in the calculation, in each kinematic bin, of the so-called deviations:

Fehler beim Parsen (MathML mit SVG- oder PNG-Rückgriff (empfohlen für moderne Browser und Barrierefreiheitswerkzeuge): Ungültige Antwort („Math extension cannot connect to Restbase.“) von Server „https://wikimedia.org/api/rest_v1/“:): {\displaystyle \Delta=\frac{A^{sin(n\phi \pm m \phi_s)}(D1)-A^{sin(n\phi \pm m \phi_s)}(D2)}{\sqrt{\sigma^2(D1)+\sigma^2(D2)}}} ,

were Fehler beim Parsen (MathML mit SVG- oder PNG-Rückgriff (empfohlen für moderne Browser und Barrierefreiheitswerkzeuge): Ungültige Antwort („Math extension cannot connect to Restbase.“) von Server „https://wikimedia.org/api/rest_v1/“:): {\displaystyle \sigma} s represent the errors. According to this forumla, deviations of Fehler beim Parsen (Konvertierungsfehler. Der Server („https://wikimedia.org/api/rest_“) hat berichtet: „Cannot get mml. Server problem.“): {\displaystyle _\pm 1_} correspond to a discrepancy of 1 standard deviation. If the deviations in the various bins fluctuate around zero without showing any systematic shift, one can argue that possible incompatibilities on the data samples do not influence the final results.

However, the final results (e.g. azimuthal moments) are usually obtained from non-trivial combinations of the bare experimental data and, in principle, can result in a partial or total compensation of possible incompatibilities in the original data. Therefore, in order to check whether two data samples are actually compatible or not, one should also look at more basic quantities. For instance, the analyzer should compare the distributions of a selection of kinematic variables (e.g. Fehler beim Parsen (MathML mit SVG- oder PNG-Rückgriff (empfohlen für moderne Browser und Barrierefreiheitswerkzeuge): Ungültige Antwort („Math extension cannot connect to Restbase.“) von Server „https://wikimedia.org/api/rest_v1/“:): {\displaystyle x} , Fehler beim Parsen (MathML mit SVG- oder PNG-Rückgriff (empfohlen für moderne Browser und Barrierefreiheitswerkzeuge): Ungültige Antwort („Math extension cannot connect to Restbase.“) von Server „https://wikimedia.org/api/rest_v1/“:): {\displaystyle y} , Fehler beim Parsen (MathML mit SVG- oder PNG-Rückgriff (empfohlen für moderne Browser und Barrierefreiheitswerkzeuge): Ungültige Antwort („Math extension cannot connect to Restbase.“) von Server „https://wikimedia.org/api/rest_v1/“:): {\displaystyle z} , Fehler beim Parsen (MathML mit SVG- oder PNG-Rückgriff (empfohlen für moderne Browser und Barrierefreiheitswerkzeuge): Ungültige Antwort („Math extension cannot connect to Restbase.“) von Server „https://wikimedia.org/api/rest_v1/“:): {\displaystyle Q^2} , Fehler beim Parsen (Konvertierungsfehler. Der Server („https://wikimedia.org/api/rest_“) hat berichtet: „Cannot get mml. Server problem.“): {\displaystyle P_{h,\perp }} , Fehler beim Parsen (MathML mit SVG- oder PNG-Rückgriff (empfohlen für moderne Browser und Barrierefreiheitswerkzeuge): Ungültige Antwort („Math extension cannot connect to Restbase.“) von Server „https://wikimedia.org/api/rest_v1/“:): {\displaystyle \theta_x} , Fehler beim Parsen (MathML mit SVG- oder PNG-Rückgriff (empfohlen für moderne Browser und Barrierefreiheitswerkzeuge): Ungültige Antwort („Math extension cannot connect to Restbase.“) von Server „https://wikimedia.org/api/rest_v1/“:): {\displaystyle \theta_y} , Fehler beim Parsen (MathML mit SVG- oder PNG-Rückgriff (empfohlen für moderne Browser und Barrierefreiheitswerkzeuge): Ungültige Antwort („Math extension cannot connect to Restbase.“) von Server „https://wikimedia.org/api/rest_v1/“:): {\displaystyle P_{lep}} and Fehler beim Parsen (MathML mit SVG- oder PNG-Rückgriff (empfohlen für moderne Browser und Barrierefreiheitswerkzeuge): Ungültige Antwort („Math extension cannot connect to Restbase.“) von Server „https://wikimedia.org/api/rest_v1/“:): {\displaystyle P_{had}} ). Since, in general, the statistics of two different data samples is different, it is convenient to normalize the two distributions for a given kinematic variable and plot their ratio. Significant deviations of the ratio from 1 would indicate possible incompatibilities between the two data samples.

Known data compatibility issues

Here some examples of data compatibility problems in specific analyses are given to both provide examples of the studies and solutions used in past analyses and to warn about specific issues that may effect your analysis. This list is not complete but will hopefully grow over time.


Problem with 2006 electron data in Fehler beim Parsen (Konvertierungsfehler. Der Server („https://wikimedia.org/api/rest_“) hat berichtet: „Cannot get mml. Server problem.“): {\displaystyle \cos(n\phi )} analysis

  • The 06H electron data is very sensitive to the binning, as can be seen here where the z range is divided into 4->5 bins.
  • The 1st moments of the raw (not unfolded) data, evaluated for groups of 10k events, shows a strange time dependence for this data set:
    • h+ h-
    • pi+ pi-
    • Only recoil magnet on low density data h+ and h-
    • Only recoil magnet on high density data h+ and h-
    • Only recoil magnet off data (only 2 points due to low statistics) h+ and h-
    • The first presentation at a cos(phi) meeting is here
    • The second presentation at a cos(phi) meeting, with the results of fitting to a linear function and the addition of MC is here
  • As a result this data set has been removed from the final results


Collins / Sivers analysis


Exclusive XXXX analysis NEEDFIX

  • Incompatibility between XX and XX data
    • See studies done by Zaven
  • Careful recalculation of the beam energy required
    • see the Beam energy page for the correction formula and details
  • for full details see the release page for details


Some other analysis.... NEEDFIX

  • Something about helicity balancing ??


Target Spring Finger Problem in 1998

  • In 1998 a spring finger, 3-4mm long and 0.5mm thick, used to connect the beam pipe to the target cell, was forgotten inside the target region, hanging from the top of the cell. Beam scattering off this finger created a peak in the zvertex distributions and affected mainly the low x region.

1998 and 1999 data

  • The data taken in 1998 and 1999 were dropped in the published g1-paper and in the F2 analysis due to the overwhelming statistics in 2000 (F2: and 2005).

Overall normalizations and top-bottom-asymmetries

  • A comparison of relative overall normalizations of DIS yields (separated into the top and bottom halves of the detector) can be found in the F2 release report

2004 unpolarized deuterium data (DVCS)


See also: Year Dependent Info 1995-2007 on the Productions page


NEEDTEXT: Review of Markus (and Rebecca) -- Add more examples of other analysis groups

Abgerufen von „https://hermes-wiki.desy.de/index.php?title=Year_Dependence&oldid=11841“