Review

Author(s): Anthony Atkinson

Review by: Anthony Atkinson
Source: Journal of the Royal Statistical Society. Series A (Statistics in Society), Vol. 158,
No. 3 (1995), pp. 630-631
Published by: Wiley for the Royal Statistical Society
Stable URL: http://www.jstor.org/stable/2983451
Accessed: 27-06-2016 07:49 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.

Wiley, Royal Statistical Society are collaborating with JSTOR to digitize, preserve and extend access to
Journal of the Royal Statistical Society. Series A (Statistics in Society)

This content downloaded from 141.211.4.224 on Mon, 27 Jun 2016 07:49:07 UTC
All use subject to http://about.jstor.org/terms
This content downloaded from 141.211.4.224 on Mon, 27 Jun 2016 07:49:07 UTC
All use subject to http://about.jstor.org/terms
 1995] REVIEWS 631

The number of references is impressive-it sometimes seems as if every paper which

includes the word outlier in the title, keywords or summary earns a mention. One
consequence is that parts of the book read like a catalogue: it is not always clear what the
authors would recommend in the analysis of data. Another consequence is that the focus of
the book feels rather narrow. The development over the last 15 years of diagnostics using
deletion and graphical methods receives rather peripheral acknowledgement. For example,
the opportunity has not been taken to use deletion methods for the generalized linear model
to tie together the material on logistic regression and the section on contingency tables with
the better-known material on regression. Structural time series models could also have been
included in this synthesis, but the time series chapter focuses exclusively on autoregressive
integrated moving average models.
Chapter 13 includes comments on the outlier handling facilities of a variety of statistical
packages, most of which disappoint the authors. It is followed by 39 tables for outlier tests.
One of the packages criticized is S-PLUS, the criticism rather missing the point that S-PLUS
is a system in which functions can be constructed for special purposes. The tests listed by
Barnett and Lewis could easily be included. But the analyses of data in Venables and Ripley
(1994), particularly Chapter 8, show how the identification and accommodation of outliers
have become just one part of an interactive computer analysis of data in which robustness,
deletion, plotting and simulation all play a part.

Reference
Venables, W. N. and Ripley, B. D. (1994) Modern Applied Statistics with S-Plus. New York: Springer.

Anthony Atkinson
London School of Economics
and Political Science

3. Reviews of United Kingdom Statistical Sources: vol. XXVII, Research and

Development. By D. L. Bosworth, R. A. Wilson and A. Young. ISBN 0 412 35640 6.
Chapman and Hall, London, 1993. 226 pp. ?65.

This volume on sources of statistics on research and development is the latest in the
discontinued series on statistical sources sponsored by the Economic and Social Research
Council and the Royal Statistical Society. When the decision was made to discontinue the
series a few volumes were in preparation-including this one-and an agreement was reached
to try to publish these remaining volumes. The commitment and perseverance of the series
editor, Michael Fleming of Loughborough University, has been critical in bringing this to
fruition.
Reviewing a book of this nature is not the most entertaining of tasks! The book is not
written to be read from cover to cover but is a source book. Its coverage is research and
development inputs (outputs are discussed in a companion volume in the series Review of UK
Intellectual Property Statistics). A substantial part of the introductory chapter and a
recurring theme throughout the volume is the definition of research and development. The
official international definition drawn from the wonderfully named Frascati Manual under
research and experimental development comprises creative work undertaken on a systematic
basis to increase the stock of knowledge, including knowledge of man, culture and society,
and the use of this stock of knowledge to devise new applications. The difficulty for the
authors in their application of this definition is its boundaries-how do providers of statistics
distinguish between research and development and other innovative activities-such as
quality testing and market research? In the university sector the distinction between post-

This content downloaded from 141.211.4.224 on Mon, 27 Jun 2016 07:49:07 UTC
All use subject to http://about.jstor.org/terms