This volume is the English version of the second edition of the bilingual textbook by Rasch, Verdooren and Gowers (1999). A parallel version in German is available fromthe same publisher.This book is intended for students and experimental scientists in all disciplines and presumes only elementary statistical knowledge. This prerequisite knowledge is summarised briefly in appendix B. Knowledge of differential and integral calculus is not necessary for the understanding of the text. Matrix notation is explained in Appendix C.As well as the correction of errors, the present edition differs from the first by the introduction of some new sections, such as that on testing the equality of two proportions (Section 3.4.4), and the inclusion of sequential tests. All new material is accompanied by descriptions of the relevant SPSS and CADEMO procedures.

Autorentext
Prof. Dr. Dr. h.c. Dieter Rasch . Von 1991 bis zur Emeritierung Professor mit Lehrstuhl für mathematische Statistik an der Universität Wageningen. Studierte Mathematik und Mathematische Statistik (Promotion 1962, Habilitation 1965). War Gründer und von 1961-1990 Leiter der Abteilung Biometrie eines Forschungszentrums und von 1973-1990 Honorarprofessor für Wahrscheinlichkeitstheorie und Mathematische Statistik an der Universität Rostock. Seit 2006 Gastprofessor an der Universität für Bodenkultur Wien. Über 260 wissenschaftliche Publikationen und 55 Fachbücher.
1949-1953 Oberschule, Zella Mehlis, Abiturabschluß
1953-1955 Student at the University of Jena, Mathematics
1955-1958 Student at the University of Leipzig, Ma thematical Statistics
1958 Master thesis on Estimation of Heritabi lities
1958-1991 Member of the Research Centre of Animal Production Dummerstorf Rostock (FZT), Germany
1960-1991 Lectures in Statistics at the University of Rostock, Germany
1961-1991 Head of the department of Biometry of the Research Centre of Animal Production Dummerstorf Rostoc
1961 PhD at the Math. Nat. Fakultät of the Universi ty of Leip zig (Factor Analysis)
1965 Habilitation (Analysis of Growth Curves) at the University of Rostock
1966-1978 Honorary Assistent Professor for Mathema tical Statistics at the University of Halle
1978-1991 Honorary Professor for Probability Theory and Mathematical Statistics at the University of Rostock
1991-2000 Professor for Mathematical Statistics, Department of Mathematics, Agricultural University, Wageningen, The Netherlands
1993-2000 Head of the Section Mathematical Statis tics
1995-1998 Head of the Department of Mathematics
since April 2000 Guest at the Department of Mathematics, Agricultural University, Wageningen, The Netherlands
2000-2004 Senior researcher, BioMath GmbH Rostock, Germany
Apri- June 2000 Guest Professor at the University of Vienna, Inst. of Statistics
June 2001 Honorary Doctor degree from the St. Izstvan University Budapest
March-July 2003 Guest Professor at the Math. Inst. of the University of Klagenfurt
Oct. 2004 Jan. 2005 Guest Professor at the Math. Inst. of the University of Klagenfurt
Since 1.9.2006 Professor at the Inst. für Statistik und EDV, Univ. Für Bodenkultur, Wien
Special research topics:
Mathematical Statistics, Biometry, Population Genetics and Breeding, Analysis of growth curves, Nonlinear Regression, Robustness, Experimental Design, Clinical Trials. Rob Verdooren: Diplom an der Landwirtschaftschaftlichen Universität Wageningen, Die
Niederlande. Promotion auf dem Gebiet der Biometrie, Emeritus "Associate Professor" für Versuchsplanung und Analyse von Versuchen Universität Wageningen. Jetzt Berater für Statistik bei Numico Research B.V. in Wageningen, Die Niederlande. J. I. Gowers, BA in Cambridge, MSc in Angewandter Statistik an der University of Bath, wissenschaftlicher Mitarbeiter für Statistik am Institute of Biometry and Community Medicine der University of Exeter, später Head of the Department of Mathematical Sciences an der University of the West of England.

Leseprobe
1 Introduction (p. 2-3)

Empirical research: we understand this term to mean the acquisition of knowledge either through passive observations (as in surveys), or from experiments, where we actively influence or manipulate the research material and then make observations. Given the steadily increasing financial problems facing research institutions, how can empirical research be conducted as efficiently (cost-effectively) as possible? This book seeks to make a contribution to answering this question.

The methods are in existence, but they are not widely used. The reasons for this stem on the one hand from the traditions of experimental research, and also from the behavioural patterns of the researchers themselves. Optimal research design implies that the objective of the investigation is determined in detail before the experiment or survey is carried out and that the precision requirements for the type of analysis planned for the data using the chosen statistical model are formulated, and that all possible things which could have a negative influence on the research work or could bias or disturb the results are considered. This all makes work, takes time, and is much more difficult than simply starting with the practical investigation and formulating the objective of the study after the data have been collected and then seeing what reliability the results have.

However, society, which in the end finances research, is becoming less and less tolerant of such ways of proceeding. Indeed in very sensitive areas, such as clinical and toxicological research, in experiments with animals or research into the cloning of organisms, it is a requirement to have the detailed research designs judged by regulatory committees before the study commences. Only after approval can the research begin. As well as the question of safeguards for mankind and nature, there is a related problem of saving costs.

Surveys and experiments are distinguished by the role which the investigators play. In surveys they hardly interfere with events, they merely observe. This is typical in economic and sociological research, but also occurs in parts of forestry science and in ecology and population biology. In other areas experiments take place, different feeds and fertilising agents are used, different therapies are compared. In industrial research one must frequently investigate the influence of various factors on the quality or quantity of an end product, or adjust production conditions to achieve an optimum. This can require the systematic variation of relevant factors.

In all these areas there are basic principles for selecting efficient (optimal) designs, which we will discuss in this volume. The most important factor here is that the design of the empirical investigation should be properly formulated. It is certainly true, and this will be clarified in examples, that this is often not a simple task, and can take up days or months. Since an experiment often lasts far longer than its analysis, one can quickly alter a false analysis of correctly obtained data, but often the most sophisticated analysis cannot save an insufficiently thought through experiment, which must therefore be repeated with consequent greatly increased costs.

Acquiring knowledge by empirical research starts with a deductive phase (preexperimental). In addition to an outline description of the research problem, this phase includes a precise elaboration of the aim of the research, the exact definition of the precision demanded in connection with, for example, the probabilities of errors of the first and second kind in hypothesis tests, or the confidence coefficient and the expected width of confidence intervals, and will also include the selection or construction of an optimal design for the experiment.

Inhalt
1;Contents;8
2;Preface;12
3;1 Introduction;13
4;2 Planning Experiments and Su…