Werner Stahel – författare

This IMA Volume in Mathematics and its Applications DIRECTIONS IN ROBUST STATISTICS AND DIAGNOSTICS is based on the proceedings of the first four weeks of the six week IMA 1989 summer program "Robustness, Diagnostics, Computing and Graphics in Statistics". An important objective of the organizers was to draw a broad set of statisticians working in robustness or diagnostics into collaboration on the challenging problems in these areas, particularly on the interface between them. We thank the organizers of the robustness and diagnostics program Noel Cressie, Thomas P. Hettmansperger, Peter J. Huber, R. Douglas Martin, and especially Werner Stahel and Sanford Weisberg who edited the proceedings. A vner Friedman Willard Miller, Jr. PREFACE Central themes of all statistics are estimation, prediction, and making decisions under uncertainty. A standard approach to these goals is through parametric mod­ elling. Parametric models can give a problem sufficient structure to allow standard, well understood paradigms to be applied to make the required inferences. If, how­ ever, the parametric model is not completely correct, then the standard inferential methods may not give reasonable answers. In the last quarter century, particularly with the advent of readily available computing, more attention has been paid to the problem of inference when the parametric model used is not correctly specified.

This IMA Volume in Mathematics and its Applications DIRECTIONS IN ROBUST STATISTICS AND DIAGNOSTICS is based on the proceedings of the first four weeks of the six week IMA 1989 summer program "Robustness, Diagnostics, Computing and Graphics in Statistics". An important objective of the organizers was to draw a broad set of statisticians working in robustness or diagnostics into collaboration on the challenging problems in these areas, particularly on the interface between them. We thank the organizers of the robustness and diagnostics program Noel Cressie, Thomas P. Hettmansperger, Peter J. Huber, R. Douglas Martin, and especially Werner Stahel and Sanford Weisberg who edited the proceedings. A vner Friedman Willard Miller, Jr. PREFACE Central themes of all statistics are estimation, prediction, and making decisions under uncertainty. A standard approach to these goals is through parametric mod­ elling. Parametric models can give a problem sufficient structure to allow standard, well understood paradigms to be applied to make the required inferences. If, how­ ever, the parametric model is not completely correct, then the standard inferential methods may not give reasonable answers. In the last quarter century, particularly with the advent of readily available computing, more attention has been paid to the problem of inference when the parametric model used is not correctly specified.

This IMA Volume in Mathematics and its Applications DIRECTIONS IN ROBUST STATISTICS AND DIAGNOSTICS is based on the proceedings of the first four weeks of the six week IMA 1989 summer program "Robustness, Diagnostics, Computing and Graphics in Statistics". An important objective of the organizers was to draw a broad set of statisticians working in robustness or diagnostics into collaboration on the challenging problems in these areas, particularly on the interface between them. We thank the organizers of the robustness and diagnostics program Noel Cressie, Thomas P. Hettmansperger, Peter J. Huber, R. Douglas Martin, and especially Werner Stahel and Sanford Weisberg who edited the proceedings. A vner Friedman Willard Miller, Jr. PREFACE Central themes of all statistics are estimation, prediction, and making decisions under uncertainty. A standard approach to these goals is through parametric mod­ elling. Parametric models can give a problem sufficient structure to allow standard, well understood paradigms to be applied to make the required inferences. If, how­ ever, the parametric model is not completely correct, then the standard inferential methods may not give reasonable answers. In the last quarter century, particularly with the advent of readily available computing, more attention has been paid to the problem of inference when the parametric model used is not correctly specified.

This IMA Volume in Mathematics and its Applications DIRECTIONS IN ROBUST STATISTICS AND DIAGNOSTICS is based on the proceedings of the first four weeks of the six week IMA 1989 summer program "Robustness, Diagnostics, Computing and Graphics in Statistics". An important objective of the organizers was to draw a broad set of statisticians working in robustness or diagnostics into collaboration on the challenging problems in these areas, particularly on the interface between them. We thank the organizers of the robustness and diagnostics program Noel Cressie, Thomas P. Hettmansperger, Peter J. Huber, R. Douglas Martin, and especially Werner Stahel and Sanford Weisberg who edited the proceedings. A vner Friedman Willard Miller, Jr. PREFACE Central themes of all statistics are estimation, prediction, and making decisions under uncertainty. A standard approach to these goals is through parametric mod­ elling. Parametric models can give a problem sufficient structure to allow standard, well understood paradigms to be applied to make the required inferences. If, how­ ever, the parametric model is not completely correct, then the standard inferential methods may not give reasonable answers. In the last quarter century, particularly with the advent of readily available computing, more attention has been paid to the problem of inference when the parametric model used is not correctly specified.

Statistics is strongly tied to applications in different scientific disciplines, and the most challenging statistical problems arise from problems in the sciences. In fact, the most innovative statistical research flows from the needs of applications in diverse settings. This volume is a testimony to the crucial role that statistics plays in scientific disciplines such as genetics and environmental sciences, among others. The articles in this volume range from human and agricultural genetic DNA research to carcinogens and chemical concentrations in the environment and to space debris and atmospheric chemistry. Also included are some articles on statistical methods which are sufficiently general and flexible to be applied to many practical situations. The papers were refereed by a panel of experts and the editors of the volume. The contributions are based on the talks presented at the Workshop on Statistics and the Sciences, held at the Centro Stefano Franscini in Ascona, Switzerland, during the week of May 23 to 28, 1999. The meeting was jointly organized by the Swiss Federal Institutes of Technology in Lausanne and Zurich, with the financial support of the Minerva Research Foundation. As the presentations at the workshop helped the participants recognize the po­ tential role that statistics can play in the sciences, we hope that this volume will help the reader to focus on the central role of statistics in the specific areas presented here and to extrapolate the results to further applications.

Statistics is strongly tied to applications in different scientific disciplines, and the most challenging statistical problems arise from problems in the sciences. In fact, the most innovative statistical research flows from the needs of applications in diverse settings. This volume is a testimony to the crucial role that statistics plays in scientific disciplines such as genetics and environmental sciences, among others. The articles in this volume range from human and agricultural genetic DNA research to carcinogens and chemical concentrations in the environment and to space debris and atmospheric chemistry. Also included are some articles on statistical methods which are sufficiently general and flexible to be applied to many practical situations. The papers were refereed by a panel of experts and the editors of the volume. The contributions are based on the talks presented at the Workshop on Statistics and the Sciences, held at the Centro Stefano Franscini in Ascona, Switzerland, during the week of May 23 to 28, 1999. The meeting was jointly organized by the Swiss Federal Institutes of Technology in Lausanne and Zurich, with the financial support of the Minerva Research Foundation. As the presentations at the workshop helped the participants recognize the po­ tential role that statistics can play in the sciences, we hope that this volume will help the reader to focus on the central role of statistics in the specific areas presented here and to extrapolate the results to further applications.

1.1 Was ist Statistische Datenanalyse? a Eine .,Statistik" ist im täglichen Sprachgebrauch eine Zusammenstellung von Zahlen über Be­ völkerungsgruppen, ökonomische Tätigkeiten, Krankheiten, Wetterlagen oder Umwelteinflüsse. Viele Statistiken beschreiben in einem weiten Sinn den Zustand des Staates; die Statistik als Disziplin hat von da ihren Namen. In den heutigen empirischen Wissenschaften werden zu allen denkbaren Fragestellungen nach Möglichkeit Daten gesammelt. Teilweise geschieht dies durch Beobachtung von Phänomenen und Prozessen, die ohne Einfluss von Forschenden ablaufen, teilweise werden die Phänomene und Prozesse in eigens geplanten Experimenten erzeugt und gesteuert. Die Statistik befasst sich heute neben dem Problem • Wie sollen welche Daten gewonnen werden? vor allem mit den Fragen • Wie soll man Daten beschreiben? und • Welche Schlüsse kann man aus Daten ziehen? b Betrachten wir ein paar Beispiele von Datensätzen aus empirischen Wissenschaften! Das folgende historische Beispiel einer Studie der Wirksamkeit zweier Medikamente ist in der Statistik berühmt geworden. Bei 10 Versuchspersonen wurde die durchschnittliche Schlaf­ verlängerung durch Medikament A gegenüber B gemessen (in Stunden; Quelle: Cushny and Peebles, 1905, 1. PhysioI. 32, 501; "Student", 1908, Biometrika 6, I). Die Ergebnisse waren 1.2 2.4 1.3 1.3 0.0 1.0 1.8 0.8 4.6 1.4. Ist Medikament A wirksamer als B? In 9 von 10 Fällen hat es sich als wirksamer erwiesen.

1.1 Was ist Statistische Datenanalyse? a Eine .,Statistik" ist im täglichen Sprachgebrauch eine Zusammenstellung von Zahlen über Be­ völkerungsgruppen, ökonomische Tätigkeiten, Krankheiten, Wetterlagen oder Umwelteinflüsse. Viele Statistiken beschreiben in einem weiten Sinn den Zustand des Staates; die Statistik als Disziplin hat von da ihren Namen. In den heutigen empirischen Wissenschaften werden zu allen denkbaren Fragestellungen nach Möglichkeit Daten gesammelt. Teilweise geschieht dies durch Beobachtung von Phänomenen und Prozessen, die ohne Einfluss von Forschenden ablaufen, teilweise werden die Phänomene und Prozesse in eigens geplanten Experimenten erzeugt und gesteuert. Die Statistik befasst sich heute neben dem Problem • Wie sollen welche Daten gewonnen werden? vor allem mit den Fragen • Wie soll man Daten beschreiben? und • Welche Schlüsse kann man aus Daten ziehen? b Betrachten wir ein paar Beispiele von Datensätzen aus empirischen Wissenschaften! Das folgende historische Beispiel einer Studie der Wirksamkeit zweier Medikamente ist in der Statistik berühmt geworden. Bei 10 Versuchspersonen wurde die durchschnittliche Schla/­ verlängerung durch Medikament A gegenüber B gemessen (in Stunden; Quelle: Cushny and Peebles, 1905, J. Physiol. 32, 501; "Student", 1908, Biometrika 6,1). Die Ergebnisse waren 1.2 2.4 1.3 1.3 0.0 1.0 1.8 0.8 4.6 1.4. Ist Medikament A wirksamer als B? In 9 von 10 Fällen hat es sich als wirksamer erwiesen.

This book grew out of a workshop on statistics in the sciences held on Monte Verita, Switzerland, in the spring of 1999. It offers a snapshot of the role played by statistics in genetics and in the environmental sciences. A few papers dwell on genetic topics, others deal with risk assessment, in particular involving exposure to chemicals. Pollution is addressed in a survey of problems relating to atmospheric chemistry, and in an article on space debris. The collection finally presents several contributions on modern statistical methods in the sciences. The book should be particuarly useful for statisticians who wish to be informed about the use of their methods in the sciences. They will also find a variety of open problems with explanations and solutions. On the other hand, the book does not require a high degree of expertise in statistics and can, on the whole, be read profitably by researchers in genetics and environmetrics.