• KS Test 'one vs all' on artificial events generated from half-normal distribution (\mu, \sigma) = (0,1)

    • 2.2.2007 Artificial Event 0 vs 83000 artificial events each having 200 'rapidities' Event 0 vs all plus several values of KS prob
    • 2.2.2007 Artificial Event 50000 vs 83000 artificial events each having 200 'rapidities' Event 50000 vs all
      • Both graphs and explicit KS prob values indicate that KS prob takes on just about 20 different values ! I am looking at the possible source of this.
    • 19.2. 2007 Explanation of the 'gaps' in KS prob histograms above: each event has 200 rapidities -> rdmax can have 200 different values with delta rdmax = 0.005. However, most of these 200 values are highly improbable. In fact, one can typically find around 20 different values of rdmax when comparing one event with 1000 other events, see here where several values of KS prob from above are shown together with their corresponding rdmax values . One can see how no "gaps" in rdmax values (packed close together for delta rdmax = 0.005) lead to the "gaps" in the corresponding KS prob values, which are scattered from 0.0086 to 0.963 ! So it is basically the inherent discrete structure of the step cumulative function which gets 'inflated' by the KS prob which leads to the 'gaps'.

  • Test 'one vs all' on 10 000 artificial events generated from const distribution

    Each event has a multiplicity from interval (100,300) generated from const distribution, 19.2.2007.
    • Event 0 vs all 10000
    • Event 2001 vs all 10000
    • Event 4001 vs all 10000
    • Event 6003 vs all 10000
    • Event 8004 vs all 10000
    • Event 9999 vs all 10000

  • Test 'all vs all' on real data with Root bug

    C++/Root code of KS test 30.1.2007.
    • First 500 events
    • First 2000 events
    • First 8000 events

  • Test 'all vs all' on real data bug FIXED

    C++/Root code of KS test 30.1.2007.
    • First 500 events

  • Test 'all vs all' on artificial events, bug FIXED

    10 000 events with rapidities uniformly distributed on (0,1), multiplicity varies uniformly on (100,300) with this random number C++/Root code (this version generates multiplicities on (800,1000) 5.3.2007.

    KS test A with Double_t z = rdmax * help1; (C++/Root code)

    KS test B with Double_t z = rdmax * (help1 + 0.12 + 0.11/help1); (C++/Root code)

    • KS test A: First 2000 events Compare with half-normal (0,1) rapidities and uniform (100,300) multiplicity (Miki) KS test A: First 2000 events
    • KS test B: First 2000 events Compare with half-normal (0,1) rapidities and uniform (100,300) multiplicity (Miki) KS test B: First 2000 events

  • Test 'all vs all' on droplets (from Boris), bug FIXED

    C++/Root code of KS test 26.2.2007.
    • T=1 MeV + Event statistics + data 1 000 events at 1 MeV
    • T=10 MeV + Event statistics + data 1 000 events at 10 MeV
    • T=50 MeV + Event statistics + data 1 000 events at 50 MeV
    • T=175 MeV + Event statistics + data 1 000 events at 175 MeV