
Computing and Statistical Data Analysis



Glen Cowan,
Royal Holloway, University of London,
phone: (01784) 44 3452, email: g.cowan@rhul.ac.uk
Monday 3 November postgrad
tutorial at Royal Holloway!
Here is a page of
directions on how to get to the Royal Holloway Physics
Department. We'll try to get started at 11:00 so people can
use a travelcard from Waterloo. If you show up earlier you
get coffee and donuts. Meet in or around my office, Wilson 262.
We will use part of the time to finish up with some lecture material
but mainly we will work on a miniproject that exercises a number
of the concepts that we've gone over so far this term. This involves
Monte Carlo models for the processes e+e > HZ and e+e > ZZ.
The goal will be to try to develop a test statistic to best discriminate
between the two. If you want to take an advance look at some of the code,
it is located here.
Archives: The course page for the
2002 lectures. Materials from the
2002 data analysis tutorial can be found
here.
Aims: The purpose of the course is to present the basic mathematical
tools needed for the statistical analysis of experimental data.
The methods will be practiced by writing and running short
computer programs. The course will also include several
lectures on the basics of computing in High Energy Physics.
Syllabus: A general
outline of the course topics.
Schedule: Mondays 12 to 1 and 2 to 3, starting 29 September for
8 weeks. Here is a provisional
timetable for all of the postgraduate lectures.
Books: The lectures will mainly follow
 G. Cowan, Statistical Data Analysis,
Clarendon Press, Oxford, 1998.
This book has its own
web site, which
contains various data analysis resources. Also useful are:
 R.J.Barlow, A Guide to the Use of Statistical Methods in the Physical
Sciences, John Wiley, 1989;
 W.T.Eadie et al., Statistical Methods in Experimental Physics,
NorthHolland, 1971;
 S.Brandt, Statistical and Computational Methods in Data
Analysis, Springer, New York, 1998;
 L.Lyons, Statistics for Nuclear and Particle Physics, CUP, 1986.
You can also download the sections on
probability,
statistics, and
Monte Carlo
(ps files) from the Review of Particle Physics by the
Particle Data Group.
And here is a paper on
unfolding I wrote for the
2002 Durham Statistics Conference
(ps,
pdf).
Lecture Notes (print out and bring to lecture):
Problem sheets:
Some material for the tutorials and supplementary notes:
 The directory here contains a C++
test program for computing uniformly distruted random numbers and
histogramming them with HBOOK (contains a useful sample GNUmakefile).
 A simple FORTRAN program,
make_data.f,
a file for compiling and linking it,
make_data.lnk,
and a file for looking at the output with PAW,
plot_data.kumac.
 Another simple FORTRAN program to demonstrate HBOOK routines,
test_hbook.f. It calls a subroutine
get_another_x_value.f,
and the whole thing can be compiled and linked with the script
test_hbook.lnk.
 A note on how to use gmake
(ps,
pdf); the example files
Makefile,
hello.cc,
goodbye.cc, and
goodbye.h.
 A FORTRAN program to demonstrate use of the random number
generator RANMAR,
test_ranmar.f,
and a script to compile and link,
test_ranmar.lnk.
 A
C++ program to generate random numbers using a simple MLCG
(GNUmakefile works for linux).
 Routines for FORTRANcallable fitting with MINUIT:
fit_galileo.f,
fcn_galileo.f,
fit_galileo.lnk,
galileo.dat.
 The same example in C++ can be found in the files
here (makefile works for linux).
 Routines for interactive fitting with MINUIT:
fitter.f,
fcn.f,
set_plot_options.f,
show_par.f,
write_line.f,
ps_output.f,
fitter.lnk,
ls_data.dat,
ls_fit.min.
A description and sample session of this program is given
here,
and there is also a plot of the fit:
ps,
pdf.
 Routines for fitting with MINUIT from PAW with vec/fit:
vecfit.kumac,
fcn_vecfit.f.
Here is some sample data,
vecfit_data.dat,
and a plot of the fit.
 Materials from the tutorial 13.12.99 at Royal Holloway:
 The program rambo.f
by Kleiss, Stirling and Ellis to generate events uniformly distributed
in phase space (needed for the test routines below).
 A test routine for rowwise ntuples
test_rwn.f, and
linking script
test_rwn.lnk.
Look at the ntuple with PAW or read it in with
read_rwn.f,
read_rwn.lnk.
You may want to consult the
ntuple section of the HBOOK manual.
 A test routine for columnwise ntuples
test_cwn.f, and
linking script
test_cwn.lnk.
Look at the ntuple with PAW or read it in with
read_cwn.f,
read_cwn.lnk.
 The program
generate_decays.f
for generating 3body decays according to a chosen matrix element.
Needs
m2_0_minus.f,
m2_1_minus.f,
rambo.f,
generate_decays.lnk.
 The program
decay_anl.f
for analyzing the ntuples produced by generate_decays.f. Needs
read_cwn_setup.f,
m2_0_minus.f,
m2_1_minus.f,
rambo.f,
decay_anl.lnk.
 The program fit_dalitz for fitting the
relative contributions of 3body decays of 0 and 1 mesons.
The main program is fitter.f,
the same as above. Needs
fcn_dalitz.f,
read_cwn_setup.f,
m2_0_minus.f,
m2_1_minus.f,
set_plot_options.f,
show_par.f,
write_line.f,
ps_output.f,
fit_dalitz.lnk,
dalitz.min.
 An ntuple
decay_data.hbook
containing the momentum vectors from an unknown mixture of 1000
0 and 1 meson decays.
 Information on the neural network program JETNET:
The online manual,
and the code for
JETNET version 3.1.
Here is a short test routine
test_jetnet.f,
which needs
generate_xy.f,
test_jetnet.lnk.
Some HBOOK histograms generated using 2 input nodes, 2 layers and 3 hidden
nodes are in
test_jetnet.his.
 Materials from the data analysis
tutorial on 5 December 2001 at UCL.
Computing links:
Glen Cowan