Name: Interdisciplinary Cluster Workshop on Statistics 2014
Start: 2014-02-17T09:00:00+01:00
End: 2014-02-18T18:00:00+01:00
Location: MPE Garching

Interdisciplinary Cluster Workshop on Statistics 2014

from Monday, 17 February 2014 (09:00) to Tuesday, 18 February 2014 (18:00)

Monday, 17 February 2014

09:00 Cosmology

09:00 - 14:15
Room: New seminar room
Statistical Modeling for Astronomy - Jogesh Babu (PSU)
Statistical Modeling for Astronomy
- Jogesh Babu (PSU)
09:00 - 10:00
Room: New seminar room Complicated models from astrophysical theory are often fit to observational data. 'Chi-square minimization' that is commonly used for fitting functions often disregards mathematical assumptions. Alternative maximum likelihood approaches are discussed. The Bayesian Information Criterion, and related penalized likelihood measures, for evaluation of model complexity are also presented. Model goodness-of-fit is commonly tested with the Kolmogorov-Smirnov (KS) test. But this test is misused in astronomy when the model parameters are estimated from the dataset under study. Recent extensions of bootstrap resampling method, a simple Monte Carlo procedure on data, are useful in estimating the null distributions in such cases including multivariate problems. The KS is also inefficient at detecting deviations between the data and model at the tails of the distribution (the Anderson-Darling test is preferred), and it cannot justifiably be applied to multivariate data. The bootstrap methods also address the 'model misspecification' problem, inference when the data are drawn from an unknown distribution which may or may not belong to a specified family of distributions.
10:00 Coffee break
Coffee break
10:00 - 10:30
Room: New seminar room
10:30 Statistical analysis of astronomical datacubes - Jogesh Babu
Statistical analysis of astronomical datacubes
- Jogesh Babu
10:30 - 11:15
Room: New seminar room With large scale panchromatic synoptic surveys becoming more common, the most common scientific product of these new generation of telescopes is a multi-dimensional "datacube" giving brightness as a function of location in the sky at many channels of frequency. The data analysis task in these datasets faces a number of serious challenges: reducing false discoveries in the presence of non-Gaussian and heteroscedastic (varying across the cube) noise; and uncovering extended astronomical structures with low surface brightness. A straight co-addition washes out the transients in the data. A robust object detection technique capable of detecting faint sources, even those not seen at all epochs which will normally be smoothed out in traditional methods, is described. The analysis at each pixel level is based on a formula similar to "Mahalanobis distance".
11:15 Large Scale Bayesian Inference in Cosmology - Jens Jasche (IAP France)
Large Scale Bayesian Inference in Cosmology
- Jens Jasche (IAP France)
11:15 - 11:45
Room: New seminar room Already the last decade has witnessed unprecedented progress in the collection of cosmological data. Presently proposed and designed future cosmological probes and surveys permit us to anticipate the upcoming avalanche of cosmological information during the next decades. The increase of valuable observations needs to be accompanied with the development of efficient and accurate information processing technology in order to analyse and interpret this data. In particular, cosmography projects, aiming at studying the origin and inhomogeneous evolution of the Universe, involve high dimensional inference methods. For example, 3d cosmological density and velocity field inference requires to explore on the order of 10^7 or more parameters. Consequently, such projects critically rely on state-of-the-art information processing techniques and, nevertheless, are often on the verge of numerical feasibility with present day computational resources. For this reason, in this talk Jens will address the problem of high dimensional Bayesian inference from cosmological data sets, subject to a variety of statistical and systematic uncertainties. In particular, he will focus on the discussion of selected Markov Chain Monte Carlo techniques, permitting to efficiently solve inference problems with on the order of 10^7 parameters. Furthermore, these methods will be exemplified in various cosmological applications, raging from 3d non-linear density and photometric redshift inference to 4d dynamic state inference. These techniques permit us to exploit cosmologically relevant information from observations to unprecedented detail and provide new insights into the origin and evolution of our Universe.
11:45 Lunch break
Lunch break
11:45 - 13:00
Room: New seminar room
13:00 The clustering of galaxies in BOSS - Ariel Sánchez (MPE)
The clustering of galaxies in BOSS
- Ariel Sánchez (MPE)
13:00 - 13:30
Room: New seminar room Driven by the potential of large-scale structure (LSS) observations to shed light on the physics behind the accelerated expansion of the Universe, several ground-breaking galaxy surveys are currently under way. The ongoing Baryon Oscillation Spectroscopic Survey (BOSS) is an example of these new surveys. By probing larger volumes than any of their predecessors, these surveys will measure the LSS of the Universe with unprecedented precision, providing new insights not only on the origin of cosmic acceleration, but also on many other important physical parameters. However, the precision demanded in these measurements requires a careful control of the potential systematic errors associated with our analysis techniques. In this talk Ariel will review the cosmological implications of the large-scale galaxy clustering in BOSS, with an emphasis on the statistical methods used to extract this information.
13:30 Bayesian model selection in spectral analysis of X-ray data - Johannes Buchner
Bayesian model selection in spectral analysis of X-ray data
- Johannes Buchner
13:30 - 14:00
Room: New seminar room X-ray data from distant sources are traditionally analysed with Maximum Likelihood methods ("fitting"), taking into account the Poisson nature of the data collection process. Johannes will present the step going to a Bayesian framework, including uncertainties e.g. from photo-metric redshifts. This allows not just parameter estimation but also effective comparison of various models, and thus physical processes. Questions such as "Is there a feature in this data?" and "How many spectral lines are there?", but also "Is it X or Y or Z that is going on here?" become tangible. An application to high-redshift AGN is discussed, where multiple geometric models of the obscuring region ("torus") are distinguished using X-ray spectra. If time permits, methods for discovering new models are discussed.
14:00 DISCUSSION
DISCUSSION
14:00 - 14:15
Room: New seminar room
14:15 Astrophysical data analys with information field theory - Torsten Ensslin (MPA)
Astrophysical data analys with information field theory
- Torsten Ensslin (MPA)
14:15 - 15:15
Room: New seminar room The problem of reconstructing an image or a function from data is generally ill-posed. The desired signal has an infinite number of degrees of freedom whereas the data is only providing a finite number of constraints. Additional statistical information and other knowledge has to be used to regularize the problem. Information field theory permits us to formulate signal inference problems rigorously using probabilistic language to combine data and knowledge. It helps us to exploit existing methods developed for field theories to derive optimal reconstruction algorithms. In this lecture, an introduction to the basic principles of information field theory will be given and illustrate by concrete examples from astrophysical applications.
Image analysis

14:15 - 17:30
Room: New seminar room
15:15 Bayesian mixture modelling - Fabrizia Guglielmetti (MPE)
Bayesian mixture modelling
- Fabrizia Guglielmetti (MPE)
15:15 - 15:45
Room: New seminar room Bayesian probability theory is combined with a two-component mixture model to analyse astronomical images. The Bayesian mixture models technique [1] is originally designed to address current astrophysical problems in high-energy astrophysics (e.g., stellar evolution, evolution of galaxies and the large-scale structure of the universe). Under this environment, images are characterised by Poisson statistics. The technique correctly handles the statistics and provides for a joint background estimation and source detection and a proper propagation of errors in background and source estimates. The technique provides for a reliable detection of background as well as sources with their respective uncertainties, improving faint source detection with respect to previous analyses. It is particularly suited to the identification of objects featuring extended emission like clusters of galaxies. Different statistics are easily adapted to the technique allowing for a wider application in science. References: [1] F. Guglielmetti et al. MNRAS 396, 165 (2009). Key Words: Methods: data analysis, statistical, Techniques: image processing
15:45 Coffee break
Coffee break
15:45 - 16:15
Room: New seminar room
16:15 The NIFTY approach to high energy photon imaging - Marco Selig (MPA)
The NIFTY approach to high energy photon imaging
- Marco Selig (MPA)
16:15 - 16:45
Room: New seminar room In this talk, NIFTY for the development of signal reconstruction algorithms that operate independent of any underlying spatial grid and its resolution is presented (NIFTY: Numerical Information Field Theory, project homepage at http://www.mpa-garching.mpg.de/ift/nifty ). A large number of Bayesian methods for 1D signal reconstruction, 2D imaging, as well as 3D tomography, appear formally similar, but one often finds individualized implementations that are neither flexible nor easily transferable. Signal inference in the framework of NIFTY can be done in an abstract way, such that algorithms, prototyped in 1D, can be applied to real world problems in higher-dimensional settings. A number of 1D, 2D, 3D and spherical examples of NIFTY's versatile application areas that involve astrophysical, medical, and abstract inference problems are presented. Those include a recent application of the D3PO algorithm targeting the non-trivial task of Denoising, Deconvolving, and Decomposing Photon Observations in high-energy astronomy.
16:45 Maximum entropy imaging in astronomy - Andy Strong (MPE)
Maximum entropy imaging in astronomy
- Andy Strong (MPE)
16:45 - 17:15
Room: New seminar room The maximum entropy method applied to imaging was first introduced in the mid-1970's by radio astronomers for interferometric data. The rationale behind the method evolved considerable in the following years, until it was placed firmly in a Bayesian framework. Andy will review the historical development and the formalism, give examples of applications in high-energy astronomy and elsewhere, and discuss what could be the role of this method in the future.
17:15 DISCUSSION
DISCUSSION
17:15 - 17:30
Room: New seminar room
Tuesday, 18 February 2014

09:00 Particle physics

09:00 - 15:15
Room: New seminar room
Statistical methods in particle physics - Glen Cowan (CERN London U)
Statistical methods in particle physics
- Glen Cowan (CERN London U)
09:00 - 10:00
Room: New seminar room The talk will review the important statistical methods in Particle Physics, with particular emphasis on those currently used at the Large Hadron Collider. This will cover statistical tests for discovering new phenomena as well as placing limits on parameters of new models. Both frequentist and Bayesian methods will be described, with particular emphasis on treatment of systematic uncertainties through introduction of nuisance parameters. The talk will also address important statistical tools such as multivariate methods for event selection and unfolding (deconvolution) of distributions.
10:00 Coffee break
Coffee break
10:00 - 10:15
Room: New seminar room
10:15 Statistical techniques for the Higgs searches at the LHC - Sandra Kortner (MPP)
Statistical techniques for the Higgs searches at the LHC
- Sandra Kortner (MPP)
10:15 - 10:45
Room: New seminar room The cross section for the production of a Higgs boson at the LHC is orders of magnitude smaller than the cross sections of competing processes like QCD jet production or the production of electroweak gauge bosons. The small Higgs cross section and the necessity to suppress the competing processes leads to a very small number of events after all selection cuts. In order to prove the existence of the Higgs boson and to measure its properties dedicated statistical methods which take into account the low statistics and the presence of systematic errors are applied in the LHC experiments. These methods combine analytic asymptotic approaches with Monte-Carlo techniques and will be summarized in the presentation.
10:45 Partial-wave techniques in hadron physics - Boris Grube (TUM)
Partial-wave techniques in hadron physics
- Boris Grube (TUM)
10:45 - 11:15
Room: New seminar room The study of the excitation spectra of baryons and mesons is an active field of hadron physics, where one wants to learn more about how hadrons are composed of quarks and gluons and how confinement works. In typical experiments many different of these excited hadrons are produced at the same time. They appear as short-lived resonances that decay into final-state particles measured by the detector. Partial-wave analysis methods are used in order to disentangle the observed mixture of the contributing resonances, by decomposing the measured kinematic distributions of the decay particles into partial waves with well-defined spin and parity quantum numbers. This spin-parity decomposition is an expansion in an orthogonal set of functions, not unlike a Fourier analysis. The coefficients of the expansion describe the strengths and phases of each partial wave and are determined by a multi-dimensional fit to the measured kinematic distributions which typically contains about 100 free parameters. One of the challenges in this kind of analysis is model selection, that is to decide how to truncate the partial-wave series. A second challenge is the extraction of resonance parameters like mass and width (life time) from the partial-wave-decomposed data. This is done by fitting models that contain a number of resonances to the data. These fits typically contain several hundred free parameters.
11:15 Cocktail of sampling methods - Frederik Beaujean (LMU, C2PAP)
Cocktail of sampling methods
- Frederik Beaujean (LMU, C2PAP)
11:15 - 11:45
Room: New seminar room Frederik will give a rough overview of Monte Carlo sampling algorithms. Special emphasis is not on how each algorithm works but rather on which kind of problems it can solve well in terms of dimensionality, multimodality, and parallel execution. If time permits, he will present some new ideas to blend basic algorithms in order to increase robustness.
11:45 Lunch break: Bavarian Brotzeit
Lunch break: Bavarian Brotzeit
11:45 - 14:00
Room: New seminar room
14:00 Data analysis for low-temperature detectors - Achim Guetlein (TUM)
Data analysis for low-temperature detectors
- Achim Guetlein (TUM)
14:00 - 14:30
Room: New seminar room Low-temperature detectors are used in several rare-event searches like direct dark matter detection or neutrinoless double beta decay. In this talk the data-analysis procedure from recorded data to final spectra will be presented. The emphasis will be on pulse shape analysis as well as the determination of cut efficiencies.
14:30 The search for neutrinoless double beta decay with GERDA: Extraction of stringent half-life limits from a few events - Bela Majorovits (MPP)
The search for neutrinoless double beta decay with GERDA: Extraction of stringent half-life limits from a few events
- Bela Majorovits (MPP)
14:30 - 15:00
Room: New seminar room A short motivation of the search for the lepton number violating neutrinoless double beta decay (0vbb) will be given. The GERDA experiment - designed and built to search for this extremely rare decay - will be introduced and the data used for 0vbb decay analysis will be shown. The analysis leading to the lower limit for 0vbb-decay in 76-Ge of T_1/2 > 2.1 *10^{25}y will be explained in some detail.
15:00 DISCUSSION
DISCUSSION
15:00 - 15:15
Room: New seminar room

Choose timezone