Statistics and Forensic Science—A Fruitful Partnership

Statistics and Forensic Science—A Fruitful Partnership

EDITORIAL Statistics and Forensic Science-A Partnership Fruitful C. G. G. AITKEN Department of Statistics, Universio of Edinburgh, Mayjeld Road, Ed...

286KB Sizes 2 Downloads 92 Views

EDITORIAL

Statistics and Forensic Science-A Partnership

Fruitful

C. G. G. AITKEN Department of Statistics, Universio of Edinburgh, Mayjeld Road, Edinburgh, United Kingdom EH9 3JZ Statisticians come in for continual criticism when they try to apply their methods in other fields of study. "Statistics can prove anything" and "lies, damned lies and statistics" are two comments which spring readily to mind and with which I am often accosted when I inform someone that I am a statistican. Forensic science is a fairly recent arrival in the field of statistical applications. Most of the statistical analyses are carried out by forensic scientists with only a smattering of statistical knowledge while very few statisticians as yet carry out analyses on forensic science data. One related aspect, though not directly concerned with forensic science, is the controversy surrounding statistics and the law. Lawyers are afraid that statisticians will usurp the role of the judge and jury, sitting as assessors in the courts, weighing up the probabilities and pronouncing a verdict on the basis of their findings. There is much evidence which is not amenable to statistical analysis and never will be. However, there is also much evidence which is suitable and it would be a pity if a lawyer's fear of statistics meant that such evidence was not produced in court. Returning to the main theme, too often statisticians are seen as enemies of the scientist, ready to do battle and attack all those who dare meddle in their subject and misuse it in so doing. This is far from the truth. Statisticians should be seen, and would like to be seen, as partners of scientists in general and forensic scientists in particular, advising the scientist a t each stage of his investigation: in the design of a survey, in the analysis and interpretation of a matching problem, for example, and in the presentation of results for publication or as evidence in a court of law. In this context, a little learning is a dangerous thing. Too often forensic scientists, in common with other scientists, have been exposed to a small dose of statistics in the form of a self-contained service course. This course is probably a neatly packaged collection of statistical techniques, mostly dealt with superficially, and the forensic scientist may well leave the course believing he is ready to tackle anything which the numerical world may throw a t him. However, the real numerical world is not like the ideal world of a service course and, the caveats given in the course having been all too soon forgotten, there will often be hidden traps of which the forensic scientist will not be aware. With no statistical knowledge he would automatically have sought statistical advice, with a little knowledge he will try to do it himself and all too often he will come to grief. This is not a plea for the abolition of service courses but rather the opposite, the elevation of statistics to a more important place in the curriculum of a scientist's training. Unfortunately, in the current financial crisis facing university education this ideal is still a long way from realisation. One of the main concerns of statistics is the measurement of variation. I n forensic science this will occur naturally, for example in the size distribution of particles within the silt fraction of soils or in the refractive indices of particles of glass. An ability to assess this variation correctly may be vital in the administration of justice. Too often have I seen, for example, confidence intervals wrongly constructed or wrongly interpreted. I t is only right that this task of

assessing variation should be left to those best equipped to do it, namely the statisticians. I would not expect to be permitted to carry out a forensic analysis; why should forensic scientists object when a statistician asks to do the statistical analysis ? Another of my fears concerns the reporting of the applications of statistics to forensic science. Most of these applications are reported in forensic science journals and the analyses have been done by forensic scientists. In my experience these analyses are of a low level or on an "ad-hoc" basis. There are few forensic science problems discussed in the statistical literature. Those which are discussed are, for the greater part, highly theoretical and not intelligible to most forensic scientists. I t seems to me that there should be more liaison between the two disciplines, not only in the investigation but in the publication of the results. This could take the form of joint publications; the statistical aspects could be emphasized in a statistical journal, the forensic aspects in a forensic journal. Certainly every paper with a statistical component which is submitted to a forensic science journal should be subject to refereeing by a statistician. There is an unjustified fear of statistics among non-numerate people. Their lack of numeracy is reflected in a distrust of conclusions drawn from tables of figures which, to them, are incomprehensible. I t is here that one aspect of a statistical training which is not perhaps fully appreciated is useful, namely the ability to consider data objectively and to present results clearly. Of course this aspect applies to any subject involving figures, not only forensic science. However, to further this end of objective data consideration, there are many statistical techniques, for example in the branches of discrimination, classification, similarity measures and sample surveys, which have proved useful in forensic science with its emphasis on identification and matching. The subject, I feel therefore, provides more scope for rigorous statistical analysis than is perhaps realized. This editorial began by quoting the phrase "Statistics can prove anything". This is, of course, nonsense. A properly conducted statistical argument, with its strengths and weaknesses explained and with a full understanding of the underlying assumptions, can contribute much towards a greater understanding of a set of data, forensic or otherwise. There may be lies, there may be damned lies but please, do not think of statistics, or statisticians, as being similarly dishonest. Statisticians are as concerned as everyone else in the legal profession with an honest, objective appraisal of numerical evidence and wish to help, in partnership with forensic scientists, in distinguishing knowledge, inference, opinion and emotional reaction. Please help us so to do.