Some of Glen's BaBar stuff


Glen Cowan, Royal Holloway, University of London, phone: (01784) 44 3452, e-mail: g.cowan@rhbnc.ac.uk

The preprint of our paper on b->s gamma with the recoil method, arXiv:0711.4889.

Some comments on EMC software QA at the 31 March 1999 phone meeting.

Some plots showing the power of various statistical tests, which could be applied to the BaBar software QA. I used a Gaussian reference histogram with mean 0, standard deviation 1, 10000 entries. The test histogram (1000 entries) is changed by (a) shifting the mean, (b) increasing the standard deviation, (c) adding a component with 5 times the standard deviation. The plots show the power of the test (probability to reject the alternative hypothesis at a significance level of 0.001) as a function of the magnitude of the change.

One problem with the binned Kolmogorov-Smirnov test as implemented by the HBOOK routine HDIFF is that the P-value for the null hypothesis is not uniform in [0,1], as it should be by construction, but rather is peaked closer to 1, as shown here. This is a known artifact of the binning of the histogram, and is not present if the test is done with the unbinned data. The result, which can be seen on the plots above, is that HDIFF is a less powerful test than the unbinned test done with TKOLMO.

The chi^2 tests on the algebraic moments were done with a FORTRAN routine moment_test.f.

A FINAL DRAFT paper on statistical issues in heavy flavour physics (and its latex source: gdc_durham.tex) written for the UK Phenomenology Workshop on Heavy Flavour and CP Violation.

A (hopefully final) DRAFT paper on separating upsilon(4S) decays from continuum events using a neural network by myself and Fabrizio Salvatore (and its latex source: fs_durham.tex and ps files of Fig. 1, r2.ps, Fig. 2, thrust_qcd.ps and Fig. 3, outann_12hid.ps)) written for the UK Phenomenology Workshop on Heavy Flavour and CP Violation.


Glen Cowan