Proposal List

All proposals submitted through Online Submission Form on IASC web site and through Proposal Form on ISI web site are indicated together below.

Proposer Name Proposer nationality Submission date Title of proposed IMP Short description Organizer Nationality Institution Speakers Possible collaborating Societies
Ramani S. Pilla and Jason Fine USA 07/10/2006 Functional Data Analysis: Theory and Applications Functional data analysis (or FDA) is concerned with the analysis of information on curves or functions. The unique aspect of functional data is the possibility that we can
incorporate information on the rates of change or derivatives of the curve using the slopes, curvatures, and other characteristics made available because these curves are intrinsically smooth. This session brings together researchers proposing new methodologies for studying functional data in the context of dynamical systems, mixture models, survival data and gene expression. Ramani S. Pilla (Case Western Reserve University, USA)

Jason Fine (University of Wisconsin, Madison, USA) 1. Jim Ramsay, McGill University (confirmed).
Title: Functional data analysis and Dynamic Systems.

2: Ramani S. Pilla, Case Western Reserve University and Catherine Loader, University of Auckland, NZ (confirmed).
Title: Functional data analysis for mixture models (tentative title)

3. Jason Fine, University of Wisconsin, Madison (confirmed). Title: Functional data analysis for survival data.

4. Hans Muller, UC Davis (confirmed).
Title: Functional data analysis for gene expression. ISI/Bernoulli
Fionn Murtagh Irish 04-12-2006 "Statistics and the Internet for Development, in e-Education,e-Health, and Other Fields, with particular reference to Africa" 1) Gathering statistics and data, and carrying out surveys: the role of the Internet and other telecom infrastructures.
2) ICT-supported statistics and statistical analysis in business,education and health.

Supported by IC4D, "Information and Communication Technologies for Development", web site, www.ict4d.org.uk Fionn Murtagh, Irish, Royal Holloway, University of London Speakers from the ICT4D Collective's Africa ICT4D university network:
(i) University of Dar es Salaam, Tanzania;
(ii) University of Education, Winneba, Ghana;
(iii) Universite Cheick Anta Diop, Senegal;
(iv) Universidade Eduardo Mondlane, Mozambique.
Speakers from Microsoft, Cisco.
Ismael Pena Lopez, Faculty of Law and Political Science, Universitat Oberta de Catalunya.
Others from government, non-governmental organisations, research.

Salvatore Ingrassia Italian 07/01/2007 Statistical Methods for Non Linear Latent Variable Models Statistical methods relating structural equation analysis are widely used in many applied sciences; recently such models have been considered also in the assessment of state intervention. Some case studies have pointed out that in many situations there are nonlinear relationships between variables in the model and taking into account these nonlinearities leads to an improvement of the understanding the links between such variables. In the last decade, many researches have focused their attention on different nonlinear approaches for modelling causal relationships from both a theoretical and applied point of view; many papers about this topic have been published not only in statistical domains but also in many relevant journals in areas like economics and management, behavioural and social sciences. This Invited Paper Session aims to present new results and try to compare different approaches in nonlinear latent variable modeling. In particular in the session the three speakers would consider methods based on 'Neural Networks and Latent variable models', 'Non linear PLS Path Modeling' and 'Polynomial structural equation models'; the aim is that this session could present different methodologies so that people working on such areas could share their knowledge on the subject. Salvatore Ingrassia, Universita di Catania (Italy) Vincenzo Esposito Vinzi (Dipartimento di Matematica e Statistica, Universita di Napoli, Italy), Mike Titterington, Department of Statistics, University of Glasgow, GB), Melanie Wall (Division of Biostatistics, University of Minnesota, USA)
Roland Fried German 12/01/2007 Statistical Online Monitoring Statistical online monitoring becomes more and more important. In ecology, we need to keep under surveillance regularly measured variables like temperatures and rainfall. In seismology, the interest is in extraordinary events like earthquakes and tsunamis. In medicine, we observe the vital signs of the critically ill almost continuously. In technology, we need to control e.g. drilling and chemical processes. In finance, we want to monitor stock markets. These applications are apparently quite diverse but they share the common interest of detecting patterns of change, high sampling frequencies and the often large dimension of the data as well as the need for automation. Additionally, we are often confronted with measurement problems calling for robust methods and typically there is not a simple steady state with which we could compare the measurements, as opposed to classical control charting. In conclusion, reliable monitoring techniques for complex stochastic processes and fast algorithms for finding the solutions in real time are needed. This invited paper session brings together experts reporting on the state of the art and their current research within this urgent interdisciplinary scientific field. Roland Fried (Department of Statistics, University of Dortmund, Germany) David Brillinger (Statistics Department, University of California, Berkeley, United States), Rainer Dahlhaus (Department of Mathematics, University of Heidelberg, Germany), Marianne Frisen (School of Business, Economics and Law, Goteborg University, Sweden), Marie Huskova (Department of Statistics, Charles University Prague, Czech Republic), Daniel Pena (Department of Statistics, University Carlos III de Madrid, Spain), Vladimir Spokoiny (Department of Mathematics, Humboldt University of Berlin, Germany)
Tomas Aluja-Banet Spanish 14-01-2007 Uncertainty in Statistical Matching Statistical Matching has been used in media research from the past 20 years, where the typical application has been to integrate consumption data with audience data, resorting mainly to ad-hoc solutions. It has been recently that it has attracted the interest of university researchers, giving rise to specialized sessions of congresses, like Compstat 2006, research projects funded by the EU (ESIS IST 2000 31071, DIASTASIS IST 2000 31083, DIACOFIS ISP 2000-31125), research papers and specialized books (Suzanne Raessler, Statistical Matching, a frequentist theory, practical applications and alternative bayesian approaches. Springer-Verlag, 2002, and Mauro Scanu, Marcello D'Orazio and Marco Di Zio. Statistical Matching, Theory and Practice, Wiley, 2006) At the same time the National Statistical Offices are more and more interested for combining information from multiple sources (between samples and between registries and samples) to limit the increasing cost of official surveys and to limit the respondent burden, and in this sense increasing the reliability of the collected data. Furthermore, the increasing importance of privacy and limitation of disclosure forces two or more statistical sources to lack of unit identifiers, and this is another motivation for the application of statistical matching procedures. Also it is used to obtain fused data sets for micro economic simulation, as for the Social Policy Simulation Database of the Statistics Canada, and the joint analysis of household income and expenditures (ISTAT). Finally, the increasing adoption of "split questionnaire survey designs" (e.g. the multipurpose survey conducted by Statistics Netherlands), inevitably should resort to statistical matching techniques when analyzing two or more variables investigated in distinct subquestionnaires. (Currently, there is a proposal of creation of a CENEX (Centre of Excellence) about Data Integration, covering Statistical Matching, led by ISTAT, Statistics Finland, Swiss Federal Statistical Office, Statistics Nederlands, INE, Statistics Austria, Czech NSI). Also, internet is another source of possible applications allowing to combine web data with survey data.Statistical Matching is a multidisciplinary problem, assembling technological informatic problems with advanced statistical estimation problems and it has many open problems yet, in particular the analysis of the uncertainty due to the lack of joint information, where it deserves further research. Mauro Scanu (Italy, ISTAT-Statistics Italy) and Tomas Aluja-Banet (Spain, Technical University of Catalonia) Suzanne Raessler (Germany, Erlangen-Nurnberg University), Christopher Moriarity (USA, George Washington University), Fritz Scheuren (USA, University of Chicago), Marcello D'Orazio (Italy, ISTAT), Robbert Renssen (The Neterlands, Statistics Netherlands), Peter Rossi (USA, University of Chicago) IASS/IAOS
Murray Cameron Australian 14-01-2007 Statistical and computational challenges from new environmental sensing systems Increasingly environmental scientists are endeavouring to understand and manage whole systems by using new sensors and by integrating information from multiple sources. Examples of the new technologies include (i) Spectrometers carried on satellites or aircraft and measuring simultaneously hundreds of frequencies at a spatial resolution measured in 10's of metres and available in real-time. (See eg http://www.space.gc.ca/asc/eng/satellites/hyper_environment.asp) (ii) Networks of wireless sensors measuring cheaply a small number of parameters at a large number of locations and communicating key summaries in real-time http://research.cens.ucla.edu/portal/page?_pageid=59,43783&_dad=portal&_schema=PORTAL (iii) Networks of free-drifting marine sensors measuring temperature and salinity of the ocean (See the Argo web site http://www-argo.ucsd.edu/ ). Each of these poses challenges around computation, sampling and inference. Bronwyn Harch, Australia, CSIRO Mark Berman (Australia, CSIRO), Mark Hansen (USA UCLA) ISI/Bernoulli
El Mostafa QANNARI France 15-01-2007 Sensometrics and chemometrics in food industry Overview of some problems that practioners in food industry are faced with :
- characterisation of food products by means of sensory data
- characterisation of food products by means digitalized signals (spectroscopy, image analysis, ...)
- Overview of the statistical methods used in this context (multiway data analysis, biased regression, PLS regression...)
-New trends.

The speakers should tackle diffrent aspects of sensometrics and chemometrics :
* Models for the analysis of sensory data
* Anlaysis of preference data
* Analysis of time intensity curves
* Prediction from digitalized signals
* Multiway data analysis

El Mostafa QANNARI Tormod Naes - Matforsk - Norway ;
Joachim Kunert - University of Dortmund - Germany ;
Pascal Schlich - INRA - France ;
Per Brockoff - Technical University of Danemark Guest
Adalbert Wilhelm German 15-01-2007 Statistical Modeling of Multimedia Content The processing of multimedia content is a key research area in the machine learning community. Many of the proposed algorithms rely on statistical modeling and exploration. This session aims at stressing the importance of the statistical aspects and components in the process of multimedia content modeling. From a statistical point of view multimedia content is characterized by large-scale samples in a high-dimensional ferature space. This session aims in bringng together statisticians and computer scientists who are using different statistical modeling techniques to enhance the content-based image retrieval process.
The main focus is put on applications of computer-intensive methods, clustering techniques, supervised learning, bayesian inference and decision theory to the analysis of multimedia content.
The area of multimedia mining is a rather new research field and it draws from many different fields in statistics. This session is tied together on one end by sharing the same application, i.e. multimedia data, and on the other end by a distinct reference to statistical models. Adalbert Wilhelm, German, International University Bremen, Germany Simon Wilson, Trinity College Dublin, Ireland;
Jia Li, Pennsylvania State University, Pittsburgh, USA;
Zoubin Gahramani, University of Cambridge, UK;
Nando de Freitas, University of British Columbia, Canada;
William F. Szewczyk USA 15-01-207 Measures of Effectiveness for Distributed Systems Distributed computing systems, such as sensor networks, present new challenges to the statistical community. The traditional approach of crafting "optimal" methods in isolation of how the data have been processed beforehand or how they will be used subsequently is no longer tenable. In fact, at high data rates statistically suboptimal routines may be the only feasible choices available. One then faces the question of assembling the processing choices across the distributed system and assessing how well the question at hand has been answered.

This session will present the background for the Measures of Effectiveness problem as well as some solutions for proposed and existing systems.

William F. Szewczyk, USA
National Security Agency Guest
Elvan Ceyhan Turkish 15-01-2007 Spatial Statistics: Theory, Applications, and Future Directions Spatial statistics relates to the analysis of the spatial aspect of data sets. All data have, implicitly or explicitly, a spatial (and a temporal) component. The reasons why spatial statistics is of import is threefold: (i) Data that are spatially (or temporally) close are usually more similar than those that are further apart. Hence, there is spatial (and temporal) dependence for most data sets. (ii) Spatial models to explain and make inference about data structures have important implications in various fields, e.g., epidemiology, population biology, and ecology. (iii) The literature on spatial statistics is substantial, but there still is a lot to uncover and many questions to answer. More than mapping and informal inference from patterns is required in spatial data analysis, since spatial data are neither the outcome of controlled experiments, nor result from random samples. This session, thus, brings together researchers who are proposing various spatial statistical methods required to address the inherent problems in spatial data analysis; and/or are specializing in various applications of spatial statistics. Moreover, the session intends to discuss the past, present, and future of the spatial statistics. Elvan Ceyhan, Koc University L. P. Fatti (University of the Witswatersrand, Johannesburg, South Africa), Peter J. Diggle (Lancaster University, UK), Adrian Baddeley (University of Western Australia, Australia), Noel A. Cressie (The Ohio State University, USA), Martin Kulldorff (Harvard Medical School, USA), Alfred Stein (International Institute for Geo-Information Science and Earth Observation, The Netherlands), Aila Saerkkae (Chalmers University of Technology, Sweden), Dietrich Stoyan (Institute of Statistics, Germany), Brian D. Ripley (University of Oxford, UK), Alexandra M. Schmidt (Universidade Federal do Rio de Janeiro, Brazil), Koji Kurihara (Okayama University, Japan), Philip Dixon (Iowa State University, USA)
Annie Morin France 16-02-2007 Random projections and dimensionality reduction In many applications of data mining, high dimensionality and huge data sets are frequent. Random projections (RP) have recently emerged as a method for dimensionality reduction in statistics. It has been used for several years in the database community. Using the Johnson-Lindenstrauss lemma which says that if you project n points in some high dimensional spacedown to a random O(log n) dimensional plane, there is a high probability that all distances are preserved within a small relative error. The goal of the session is to provide some recent developments and applications of RP in the statistical data mining community. Annie Morin, Universite de Rennes 1 T Hastie, USA
H Mannila, Finland
S Vempala, USA

Back to Top

Proposer Name	Proposer nationality	Submission date	Title of proposed IMP	Short description	Organizer Nationality Institution	Speakers	Possible collaborating Societies
Ramani S. Pilla and Jason Fine	USA	07/10/2006	Functional Data Analysis: Theory and Applications	Functional data analysis (or FDA) is concerned with the analysis of information on curves or functions. The unique aspect of functional data is the possibility that we can incorporate information on the rates of change or derivatives of the curve using the slopes, curvatures, and other characteristics made available because these curves are intrinsically smooth. This session brings together researchers proposing new methodologies for studying functional data in the context of dynamical systems, mixture models, survival data and gene expression.	Ramani S. Pilla (Case Western Reserve University, USA) Jason Fine (University of Wisconsin, Madison, USA)	1. Jim Ramsay, McGill University (confirmed). Title: Functional data analysis and Dynamic Systems. 2: Ramani S. Pilla, Case Western Reserve University and Catherine Loader, University of Auckland, NZ (confirmed). Title: Functional data analysis for mixture models (tentative title) 3. Jason Fine, University of Wisconsin, Madison (confirmed). Title: Functional data analysis for survival data. 4. Hans Muller, UC Davis (confirmed). Title: Functional data analysis for gene expression.	ISI/Bernoulli
Fionn Murtagh	Irish	04-12-2006	"Statistics and the Internet for Development, in e-Education,e-Health, and Other Fields, with particular reference to Africa"	1) Gathering statistics and data, and carrying out surveys: the role of the Internet and other telecom infrastructures. 2) ICT-supported statistics and statistical analysis in business,education and health. Supported by IC4D, "Information and Communication Technologies for Development", web site, www.ict4d.org.uk	Fionn Murtagh, Irish, Royal Holloway, University of London	Speakers from the ICT4D Collective's Africa ICT4D university network: (i) University of Dar es Salaam, Tanzania; (ii) University of Education, Winneba, Ghana; (iii) Universite Cheick Anta Diop, Senegal; (iv) Universidade Eduardo Mondlane, Mozambique. Speakers from Microsoft, Cisco. Ismael Pena Lopez, Faculty of Law and Political Science, Universitat Oberta de Catalunya. Others from government, non-governmental organisations, research.
Salvatore Ingrassia	Italian	07/01/2007	Statistical Methods for Non Linear Latent Variable Models	Statistical methods relating structural equation analysis are widely used in many applied sciences; recently such models have been considered also in the assessment of state intervention. Some case studies have pointed out that in many situations there are nonlinear relationships between variables in the model and taking into account these nonlinearities leads to an improvement of the understanding the links between such variables. In the last decade, many researches have focused their attention on different nonlinear approaches for modelling causal relationships from both a theoretical and applied point of view; many papers about this topic have been published not only in statistical domains but also in many relevant journals in areas like economics and management, behavioural and social sciences. This Invited Paper Session aims to present new results and try to compare different approaches in nonlinear latent variable modeling. In particular in the session the three speakers would consider methods based on 'Neural Networks and Latent variable models', 'Non linear PLS Path Modeling' and 'Polynomial structural equation models'; the aim is that this session could present different methodologies so that people working on such areas could share their knowledge on the subject.	Salvatore Ingrassia, Universita di Catania (Italy)	Vincenzo Esposito Vinzi (Dipartimento di Matematica e Statistica, Universita di Napoli, Italy), Mike Titterington, Department of Statistics, University of Glasgow, GB), Melanie Wall (Division of Biostatistics, University of Minnesota, USA)
Roland Fried	German	12/01/2007	Statistical Online Monitoring	Statistical online monitoring becomes more and more important. In ecology, we need to keep under surveillance regularly measured variables like temperatures and rainfall. In seismology, the interest is in extraordinary events like earthquakes and tsunamis. In medicine, we observe the vital signs of the critically ill almost continuously. In technology, we need to control e.g. drilling and chemical processes. In finance, we want to monitor stock markets. These applications are apparently quite diverse but they share the common interest of detecting patterns of change, high sampling frequencies and the often large dimension of the data as well as the need for automation. Additionally, we are often confronted with measurement problems calling for robust methods and typically there is not a simple steady state with which we could compare the measurements, as opposed to classical control charting. In conclusion, reliable monitoring techniques for complex stochastic processes and fast algorithms for finding the solutions in real time are needed. This invited paper session brings together experts reporting on the state of the art and their current research within this urgent interdisciplinary scientific field.	Roland Fried (Department of Statistics, University of Dortmund, Germany)	David Brillinger (Statistics Department, University of California, Berkeley, United States), Rainer Dahlhaus (Department of Mathematics, University of Heidelberg, Germany), Marianne Frisen (School of Business, Economics and Law, Goteborg University, Sweden), Marie Huskova (Department of Statistics, Charles University Prague, Czech Republic), Daniel Pena (Department of Statistics, University Carlos III de Madrid, Spain), Vladimir Spokoiny (Department of Mathematics, Humboldt University of Berlin, Germany)
Tomas Aluja-Banet	Spanish	14-01-2007	Uncertainty in Statistical Matching	Statistical Matching has been used in media research from the past 20 years, where the typical application has been to integrate consumption data with audience data, resorting mainly to ad-hoc solutions. It has been recently that it has attracted the interest of university researchers, giving rise to specialized sessions of congresses, like Compstat 2006, research projects funded by the EU (ESIS IST 2000 31071, DIASTASIS IST 2000 31083, DIACOFIS ISP 2000-31125), research papers and specialized books (Suzanne Raessler, Statistical Matching, a frequentist theory, practical applications and alternative bayesian approaches. Springer-Verlag, 2002, and Mauro Scanu, Marcello D'Orazio and Marco Di Zio. Statistical Matching, Theory and Practice, Wiley, 2006) At the same time the National Statistical Offices are more and more interested for combining information from multiple sources (between samples and between registries and samples) to limit the increasing cost of official surveys and to limit the respondent burden, and in this sense increasing the reliability of the collected data. Furthermore, the increasing importance of privacy and limitation of disclosure forces two or more statistical sources to lack of unit identifiers, and this is another motivation for the application of statistical matching procedures. Also it is used to obtain fused data sets for micro economic simulation, as for the Social Policy Simulation Database of the Statistics Canada, and the joint analysis of household income and expenditures (ISTAT). Finally, the increasing adoption of "split questionnaire survey designs" (e.g. the multipurpose survey conducted by Statistics Netherlands), inevitably should resort to statistical matching techniques when analyzing two or more variables investigated in distinct subquestionnaires. (Currently, there is a proposal of creation of a CENEX (Centre of Excellence) about Data Integration, covering Statistical Matching, led by ISTAT, Statistics Finland, Swiss Federal Statistical Office, Statistics Nederlands, INE, Statistics Austria, Czech NSI). Also, internet is another source of possible applications allowing to combine web data with survey data.Statistical Matching is a multidisciplinary problem, assembling technological informatic problems with advanced statistical estimation problems and it has many open problems yet, in particular the analysis of the uncertainty due to the lack of joint information, where it deserves further research.	Mauro Scanu (Italy, ISTAT-Statistics Italy) and Tomas Aluja-Banet (Spain, Technical University of Catalonia)	Suzanne Raessler (Germany, Erlangen-Nurnberg University), Christopher Moriarity (USA, George Washington University), Fritz Scheuren (USA, University of Chicago), Marcello D'Orazio (Italy, ISTAT), Robbert Renssen (The Neterlands, Statistics Netherlands), Peter Rossi (USA, University of Chicago)	IASS/IAOS
Murray Cameron	Australian	14-01-2007	Statistical and computational challenges from new environmental sensing systems	Increasingly environmental scientists are endeavouring to understand and manage whole systems by using new sensors and by integrating information from multiple sources. Examples of the new technologies include (i) Spectrometers carried on satellites or aircraft and measuring simultaneously hundreds of frequencies at a spatial resolution measured in 10's of metres and available in real-time. (See eg http://www.space.gc.ca/asc/eng/satellites/hyper_environment.asp) (ii) Networks of wireless sensors measuring cheaply a small number of parameters at a large number of locations and communicating key summaries in real-time http://research.cens.ucla.edu/portal/page?_pageid=59,43783&_dad=portal&_schema=PORTAL (iii) Networks of free-drifting marine sensors measuring temperature and salinity of the ocean (See the Argo web site http://www-argo.ucsd.edu/ ). Each of these poses challenges around computation, sampling and inference.	Bronwyn Harch, Australia, CSIRO	Mark Berman (Australia, CSIRO), Mark Hansen (USA UCLA)	ISI/Bernoulli
El Mostafa QANNARI	France	15-01-2007	Sensometrics and chemometrics in food industry	Overview of some problems that practioners in food industry are faced with : - characterisation of food products by means of sensory data - characterisation of food products by means digitalized signals (spectroscopy, image analysis, ...) - Overview of the statistical methods used in this context (multiway data analysis, biased regression, PLS regression...) -New trends. The speakers should tackle diffrent aspects of sensometrics and chemometrics : * Models for the analysis of sensory data * Anlaysis of preference data * Analysis of time intensity curves * Prediction from digitalized signals * Multiway data analysis	El Mostafa QANNARI	Tormod Naes - Matforsk - Norway ; Joachim Kunert - University of Dortmund - Germany ; Pascal Schlich - INRA - France ; Per Brockoff - Technical University of Danemark	Guest
Adalbert Wilhelm	German	15-01-2007	Statistical Modeling of Multimedia Content	The processing of multimedia content is a key research area in the machine learning community. Many of the proposed algorithms rely on statistical modeling and exploration. This session aims at stressing the importance of the statistical aspects and components in the process of multimedia content modeling. From a statistical point of view multimedia content is characterized by large-scale samples in a high-dimensional ferature space. This session aims in bringng together statisticians and computer scientists who are using different statistical modeling techniques to enhance the content-based image retrieval process. The main focus is put on applications of computer-intensive methods, clustering techniques, supervised learning, bayesian inference and decision theory to the analysis of multimedia content. The area of multimedia mining is a rather new research field and it draws from many different fields in statistics. This session is tied together on one end by sharing the same application, i.e. multimedia data, and on the other end by a distinct reference to statistical models.	Adalbert Wilhelm, German, International University Bremen, Germany	Simon Wilson, Trinity College Dublin, Ireland; Jia Li, Pennsylvania State University, Pittsburgh, USA; Zoubin Gahramani, University of Cambridge, UK; Nando de Freitas, University of British Columbia, Canada;
William F. Szewczyk	USA	15-01-207	Measures of Effectiveness for Distributed Systems	Distributed computing systems, such as sensor networks, present new challenges to the statistical community. The traditional approach of crafting "optimal" methods in isolation of how the data have been processed beforehand or how they will be used subsequently is no longer tenable. In fact, at high data rates statistically suboptimal routines may be the only feasible choices available. One then faces the question of assembling the processing choices across the distributed system and assessing how well the question at hand has been answered. This session will present the background for the Measures of Effectiveness problem as well as some solutions for proposed and existing systems.	William F. Szewczyk, USA National Security Agency		Guest
Elvan Ceyhan	Turkish	15-01-2007	Spatial Statistics: Theory, Applications, and Future Directions	Spatial statistics relates to the analysis of the spatial aspect of data sets. All data have, implicitly or explicitly, a spatial (and a temporal) component. The reasons why spatial statistics is of import is threefold: (i) Data that are spatially (or temporally) close are usually more similar than those that are further apart. Hence, there is spatial (and temporal) dependence for most data sets. (ii) Spatial models to explain and make inference about data structures have important implications in various fields, e.g., epidemiology, population biology, and ecology. (iii) The literature on spatial statistics is substantial, but there still is a lot to uncover and many questions to answer. More than mapping and informal inference from patterns is required in spatial data analysis, since spatial data are neither the outcome of controlled experiments, nor result from random samples. This session, thus, brings together researchers who are proposing various spatial statistical methods required to address the inherent problems in spatial data analysis; and/or are specializing in various applications of spatial statistics. Moreover, the session intends to discuss the past, present, and future of the spatial statistics.	Elvan Ceyhan, Koc University	L. P. Fatti (University of the Witswatersrand, Johannesburg, South Africa), Peter J. Diggle (Lancaster University, UK), Adrian Baddeley (University of Western Australia, Australia), Noel A. Cressie (The Ohio State University, USA), Martin Kulldorff (Harvard Medical School, USA), Alfred Stein (International Institute for Geo-Information Science and Earth Observation, The Netherlands), Aila Saerkkae (Chalmers University of Technology, Sweden), Dietrich Stoyan (Institute of Statistics, Germany), Brian D. Ripley (University of Oxford, UK), Alexandra M. Schmidt (Universidade Federal do Rio de Janeiro, Brazil), Koji Kurihara (Okayama University, Japan), Philip Dixon (Iowa State University, USA)
Annie Morin	France	16-02-2007	Random projections and dimensionality reduction	In many applications of data mining, high dimensionality and huge data sets are frequent. Random projections (RP) have recently emerged as a method for dimensionality reduction in statistics. It has been used for several years in the database community. Using the Johnson-Lindenstrauss lemma which says that if you project n points in some high dimensional spacedown to a random O(log n) dimensional plane, there is a high probability that all distances are preserved within a small relative error. The goal of the session is to provide some recent developments and applications of RP in the statistical data mining community.	Annie Morin, Universite de Rennes 1	T Hastie, USA H Mannila, Finland S Vempala, USA