Next: About this document ...
Up: Bayesian Inference in Processing
Previous: Metropolis algorithm
The gain in popularity Bayesian methods have enjoyed in recent years
is due to various conceptual and practical advantages
they have over other approaches, among which are:
- the recovery of the intuitive idea of probability as a valid
concept for treating scientific problems;
- the simplicity and naturalness of the basic tool;
- the capability of combining prior knowledge and experimental
information;
- the property permitting automatic updating as soon as new
information becomes available;
- the transparency of the methods, which allow the
different assumptions upon which an inference may depend
to be checked and changed;
- the high degree of awareness the methods give to the user.
In this article we have seen how to build a theory
of uncertainty in measurement as a straightforward application
of the basic Bayesian ideas, without unnecessary
principles or ad hoc prescriptions. In particular,
the uncertainty due to systematic errors can be treated
in a consistent and powerful way.
Providing an exact solution for inferential problems
can easily lead to computational difficulties.
We have seen several ways to overcome such difficulties,
either by using suitable approximations, or by using modern
computational methods. In particular, it has been shown
that the approximate solution often coincides
with a `conventional' method, but only under well
defined conditions. Thus, for example,
minimum
formulae can be used, with a Bayesian
spirit and with a natural interpretation
of the results, in all those routine cases
in which the analyst considers as reasonable the conditions
of their validity.
A variety of examples of applications have been shown, or
mentioned, in this paper. Nevertheless, the aim of the author
was not to provide a complete review of
Bayesian methods and applications, but rather to introduce
those Bayesian ideas that could be of help in understanding
more specialized literature.
Compendia of the Bayesian theory are
given in (Bernardo and Smith 1994, O'Hagan A 1994 and
Robert 2001). Classic, influential books are
(Jeffreys 1961, de Finetti 1974, Jaynes 1998).
Among the many books introducing Bayesian methods,
(Sivia 1996) is particularly suitable for physicists.
Other recommended texts which treat general aspects of data analysis are
(Box and Tiao 1973, Bretthorst 1988,
Lee 1989, Gelman et al 1995,
Cowell et al 1999, Denison et al 2002,
Press 2002).
More specific applications can be found in the
proceedings of the conference series and several web sites.
Some useful starting points for web navigation are given:
ISBA book list |
http://www.bayesian.org/books/books.html |
UAI proceedings |
http://www2.sis.pitt.edu/ dsl/UAI/uai.html |
BIPS |
http://astrosun.tn.cornell.edu/staff/loredo/bayes/ |
BLIP |
http://www.ar-tiste.com/blip.html |
IPP Bayesian analysis group |
http://www.ipp.mpg.de/OP/Datenanalyse/ |
Valencia meetings |
http://www.uv.es/~bernardo/valenciam.html |
Maximum Entropy resources |
http://omega.albany.edu:8008/maxent.html |
MCMC preprint service |
http://www.statslab.cam.ac.uk/~mcmc/ |
I am indebted to Volker Dose and Ken Hanson for extensive
discussions concerning the contents of this paper, as well as
for substantial editorial help.
The manuscript has also benefited from comments
by Tom Loredo.
References
- Astone P et al 2002 Search for correlation
between GRB's detected by BeppoSAX
and gravitational wave detectors EXPLORER and NAUTILUS
Phys. Rev. 66 102002.
- Astone P and D'Agostini G 1999
Inferring the intensity of Poisson processes at limit
of detector sensitivity (with a case study on gravitational wave burst search)
CERN-EP/99-126
- Astone P, D'Agostini G and D'Antonio 2003
Bayesian model comparison applied to the
Explorer-Nautilus 2001 coincidence data,
arXiv:gr-qc/0304096
- Babu G J and Feigelson E D 1992 eds
Statistical Challenges in Modern Astronomy I (New York: Springer)
- Babu G J and Feigelson E D 1997 eds
Statistical Challenges in Modern Astronomy II (New York: Springer)
- Berger J O and Jefferys W H 1992
Sharpening Ockham's razor on a Bayesian strop
Am. Scientist 89 64-72
and J. Ital. Stat. Soc. 1 17
- Bernardo J M 1999 ed Bayesian Methods in the Sciences,
special issue of Rev. Acad. Cien. Madrid, 93(3)
- Bernardo J M, J O Berger, A P Dawid and A F M Smith 1999 eds
Bayesian Statistics 6 (Oxford: Oxford University)
- Bernardo J M and Smith F M 1994 Bayesian Theory
(Chichester: John Wiley & Sons)
- Bernardo J M 1997 Non-informative priors do not exist
J. Stat. Planning and Infer. 65 159
- Bontekoe T R, Koper E and Kester D J M 1994
Pyramid maximum entropy images of IRAS survey data
Astron. Astrophys. 284 1037
- Bouman C A and Sauer K 1993 A generalized Gaussian image model
for edge-preserving MAP estimation
IEEE Trans. on Image Processing 2 296
- Box G E P and Tiao G C 1973
Bayesian inference in statistical analysis
(Chichester: J. Wiley & Sons)
- Bretthorst G L 1988 Bayesian spectrum analysis and parameter
estimation (Berlin: Springer)
- BUGS 1996 http://www.mrc-bsu.cam.ac.uk/bugs/welcome.shtml
- Buck B and Macauly V A eds 1991 Maximum Entropy in action,
(Oxford: Oxford University Press)
- Ciuchini M et al 2001
2000 CKM-Triangle Analysis: A critical review
with updated experimental inputs and theoretical parameters
J. High Energy Phys. 0107 013
- Coletti G and Scozzafava R 2002
Probabilistic logic in a coherent setting'',
(Dordrecht: Kluwer Academic)
- Cousins R D 1995 Why isn't every physicist a Bayesian?
Am. J. Phys. 63 398
- Cowell R G, Dawid A P, Lauritzen S L and Spiegelhalter D J 1999
Probabilistic Networks and Expert Systems, (New York: Springer)
- Cox R T 1946 Probability, Frequency and Reasonable Expectation
Am. J. Phys. 14 1
- Cozman F B 2001
JavaBayes version 0.346 - Bayesian networks in Java
http://www-2. cs.cmu.edu/~javabayes/Home/
- Cunningham G S, Hanson K M and Battle X L 1998
Three-dimensional reconstructions from low-count SPECT data
using deformable models Opt. Expr. 2 227
- D'Agostini G 1999a Bayesian Reasoning versus Conventional Statistics
in High Energy Physics
Maximum Entropy and Bayesian Methods
ed von der Linden W et al (Dordrecht: Kluwer Academic)
- D'Agostini G 1999b Sceptical combination of experimental results:
General considerations and application to epsilon-prime/epsilon
CERN-EP/99-139
- D'Agostini G 1999c Bayesian reasoning in high-energy
physics: principles and applications CERN Report 99-03
(an extended version of this report is going to be published as
Bayesian reasoning in data analysis - A critical introduction
by World Scientific Publishing)
- D'Agostini G 1999d Teaching statistics in the physics curriculum:
Unifying and clarifying role of subjective probability
Am. J. Phys. 67 1260
- D'Agostini G 1999e Overcoming prior Anxiety,
Bayesian Methods in the Sciences ed
J. M. Bernardo;
special issue of Rev. Acad. Cien. Madrid 93(3),
311
- D'Agostini G 2000 Confidence limits: what is the problem?
Is there the solution?
CERN Report 2000-005
ed James F and Lyons L (Geneva: CERN) 3
- D'Agostini G 2002 Minimum bias legacy of search results 2002
Nucl Phys Proc Suppl 109 148
- D'Agostini G and Degrassi G 1999
Constraints on the Higgs Boson Mass from Direct
Searches and Precision Measurements
Eur. Phys. J. C10 663
- D'Agostini G and Raso M
Uncertainties due to imperfect knowledge of systematic effects:
general considerations and approximate formulae
CERN-EP/2000-026
- de Finetti B 1974 Theory of Probability
(Chichester: J. Wiley & Sons)
- Denison D G T, Holmes C C, Mallick B K and Smith A F M 2002
Bayesian methods for nonlinear classification and regression
(Cichester: J. Wiley & Sons)
- DIN (Deutsches Institut für Normung) 1996
Grundlagen der Messtechnik - Teil 3:
Auswertung von Messungen einer einzelnen Messgröße, Messunsicherheit DIN 1319-3
(Berlin: Beuth Verlag)
- DIN (Deutsches Institut für Normung) 1999
Grundlagen der Messtechnik - Teil 4: Auswertung von Messungen, Messunsicherheit
DIN 1319-4 (Berlin: Beuth Verlag)
- Dose V 2002 Bayes in five days,
CIPS, MPI für Plasmaphysik, Garching, Germany, Reprint 83, May 2002
- Dose V and von der Linden W 1999 Outlier tolerant parameter
estimation Maximum Entropy and
Bayesian Methods ed von der Linden W et al (Dordrecht: Kluwer Academic) 47
- Efron B 1986a Why isn't everyone a Bayesian?
Am. Stat. 40 1
- Efron B 1986b reply to Zellner 1986 Am. Stat. 40 331
- Fischer R, Mayer M, von der Linden W and Dose V 1997
Enhancement of the energy resolution in ion-beam
experiments with the maximum-entropy method
Phys. Rev. E 55 6667
- Fischer R, Mayer M, von der Linden W and Dose V 1998
Energy resolution enhancement in ion beam experiments
with Bayesian probability theory Nucl. Instr. Meth. 136-138
1140
- Fischer R, Hanson K M, Dose V and von der Linden W 2000
Background estimation in experimental spectra Phys. Rev. E61 1152
- Fröhner F H 2000 Evaluation and Analysis of Nuclear
Resonance Data JEFF Report 18 (Paris: OECD Publications)
- Gelman A, Carlin J B, Stern H S and Rubin D B 1995 Bayesian Data Analysis (London: Chapman and Hall)
- Glimm J and Sharp D H 1999 Prediction and the quantification of uncertainty
Physica D 133 152
- Gregory P C and Loredo T J 1992
A new method for the detection of a periodic signal
of unknown shape and period Astr. J. 398 146
- Gregory P C and Loredo T J 1996 Bayesian periodic signal detection II
- Bayesian periodic signal detection: analysis of ROSAT observations
of PSR 0540-693 Astr. J. 473 1059
- Gregory P C 1999
Bayesian periodic signal detection I -
analysis of 20 years of radio flux measurements of
the x-ray binary LS I +61
303
Astr. J. 520 361
- Gubernatis J E, Jarrell M, Silver R N and Sivia D S 1991
Quantum Monte-Carlo simulations and maximum-entropy: dynamics
from imaginary-time data Phys. Rev. B 44 6011
- Hanson K M 1993
Introduction to Bayesian image analysis
Medical Imaging: Image Processing
Loew M H ed Proc. SPIE 1898 716
- Hanson K M 2000 Tutorial on Markov Chain Monte Carlo,
http://public.lanl.gov/ kmh/talks/maxent00b.pdf
- Hasting W K 1970 Monte Carlo sampling methods using Markov chains and
their applications Biometrica 57 97
- Higdon D M and Yamamoto S Y 2001
Estimation of the head sensitivity function
in scanning magnetoresistance microscopy
J. Amer. Stat. Assoc. 96 785
- Hobson M P, Bridle S L and Lahav 2002
Combining cosmological datasets:
hyperparameters and Bayesian evidence
arXiv:astro-ph/0203259
- Howson C and Urbach P 1993
Scientific reasoning -- the Bayesian approach (Chicago and La Salle: Open Court)
- ISO (International Organization for Standardization) 1993
Guide to the Expression of Uncertainty in Measurement (Geneva: ISO)
- Jaynes E T 1957a Information theory and statistical mechanics Phys. Rev.
106 620
- Jaynes E T 1957b Information theory and statistical mechanics II
Phys. Rev. 108 171
- Jaynes E T 1968 Prior probabilities IEEE Trans. Syst. Cybern.
SSC-4 227, reprinted in (Jaynes 1983)
- Jaynes E T 1973 The well-posed problem
Found. Phys. 3 477, reprinted in (Jaynes 1983)
- Jaynes E T 1983 Papers on Probability, Statistics
and Statistical Physics ed Harper W L and Hooker C A (Dordrecht: Reidel)
- Jaynes E T 1998 http://omega.albany.edu:8008/JaynesBook.html
- Jeffreys H 1961 Theory of Probability (Oxford: Oxford University)
- John M V and Narlikar J V 2002
Comparison of cosmological models using Bayesian theory Phys. Rev.
D65 43506
- Kadane J B and Schum D A 1996 A Probabilistic Analysis of the
Sacco and Vanzetti Evidence (Chichester: Wiley and Sons)
- Kalman R E 1960 A new approach to linear filtering and prediction problems
Trans. ASME Journal of Casic Engineering 82 35
- Kass R E, Carlin B P, Gelman A and Neal R M
1998 Markov Chain Monte Carlo in practice: A roundtable discussion
Am. Stat. 52 93
- Lad F 1996 Operational Subjective Statistical Methods -
a Mathematical, Philosophical, and Historical Introduction
(Chichester:J. Wiley & Sons)
- Lee P M 1989 Bayesian statistics - an introduction
(Chichester:J. Wiley & Sons)
- Lewis A and Bridle S 2002
Cosmological parameters from CMB and other data:
a Monte-Carlo approach Phys.Rev. D66 103511
- von der Linden W 1995 Maximum-entropy data analysis
Appl. Phys. A60 155
- von der Linden W, Dose V and Fischer R 1996b
Spline-based adaptive resolution image reconstruction
Proceedings of the 1996 Maximum Entropy Conference
ed Sears M et al (Port Elizabeth: N.M.B. Printers) 154
- Lindley D V 1986 Discussion to Efron 1986a Am. Stat. 40 6
- Loredo T J 1990 Maximum Entropy and Bayesian Methods
ed Fougére P F (Dordrecht: Kluwer Academic) 81
- Loredo T J and Lamb D Q 2002
Bayesian analysis of neutrinos observed from supernova SN 1987A
Phys. Rev. D65 063002
- Malakoff D 1999 Bayes Offers a 'New' Way to Make Sense of Numbers
Science 286 1460
- Maybaeck P S 1979 Stochastic models, estimation and control,
Vol. 1 (New York: Academic Press).
- Metropolis H, Rosenbluth A W, Rosenbluth M N, Teller A H and
Teller E 1953 Equations of state calculations by fast computing
machines Journal of Chemical Physics 21 1087
- von Mises R 1957 Probability, Statistics, and Truth
(St Leonards: Allen and Unwin); reprinted in 1987 by Dover
- Neal R M 1993 Probabilistic inference using
Markov chain Monte Carlo methods (Toronto: Technical Report CRG-TR-93-1)
- O'Hagan A 1994 Kendall's Advanced Theory of
Statistics: Vol. 2B - Bayesian Inference (New York: Halsted)
- Pearl J 1988
Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference
(San Mateo: Morgan Kaufmann)
- Press W H 1997 Understanding data better with
Bayesian and global statistical methods
Unsolved problems in astrophysics49-60
ed Bahcall J N and Ostriker J P
(Princeton: Princeton University) 49
- Press S J 2002 Subjective and Objective Bayesian Statistics:
Principles, Models, and Applications 2nd Edition
(Chichester: John Wiley & Sons)
- Robert C P 2001 The Bayesian Choice (New York: Springer)
- Saquib S S, Hanson K M, and Cunningham G S 1997
Model-based image reconstruction from time-resolved diffusion data
Proc. SPIE 3034 369
- Schrödinger E 1947a The Foundation of the Theory of Probability - I
Proc. R. Irish Acad.
51A 51; reprinted in Collected papers Vol. 1
(Vienna 1984: Austrian Academy of Science) 463
- Schrödinger E 1947b The Foundation of the Theory of Probability - II
Proc. R. Irish Acad.
51A 141; reprinted in Collected papers Vol. 1
(Vienna 1984: Austrian Academy of Science) 479
- Sivia D S 1997 Data Analysis - a Bayesian Tutorial
(Oxford: Clarendon)
- Skilling J 1992 Quantified maximum entropy
Int. Spectr. Lab. 2 4
- Smith A F M 1991 Bayesian numerical analysis
Phyl. Trans. R. Soc. London 337 369
- Taylor B N and Kuyatt C E 1994
Guidelines for Evaluating and Expressing Uncertainty of
NIST Measurement Results
(Gaithersburg: NIST Technical Note 1297);
available on line at http://physics.nist.gov/
- Tribus M 1969 Rational Descriptions, Decisions, and Designs
(Elmsford: Pergamon)
- Welch G and Bishop G 2002 An introduction to Kalman filter
(http://www.cs.unc. edu/~welch/kalman/)
- Zech G 2002 Frequentist and Bayesian confidence limits
Eur. Phys. J. direct C12 1
- Zellner A 1986 Bayesian solution to a problem posed by Efron
Am. Stat. 40 330
Next: About this document ...
Up: Bayesian Inference in Processing
Previous: Metropolis algorithm
Giulio D'Agostini
2003-05-13