Statistics for Archaeologists: A Commonsense Approach. By ROBERT D. DRENNAN. Plenum Press, New York and London. 1996. xix + 278 pp., figures, tables, suggested readings, index. $42.50 (cloth ISBN 0-306-45327-4; paper ISBN 0-306453264).
I have been teaching statistics in anthropology or archaeology departments for more than a decade. Any new archaeological statistics book is therefore a welcome addition, for in them one can find new material, approaches, data sets, examples, and modes of presentation for a topic many view as complex. This can certainly be said of Statistics for Archaeologists, but as I pored over this volume the thought crossed my mind that perhaps I have been at it too long, for much of this material seemed mundane. All too frequently excessive detail and descriptions are presented, making some sections suitable only for readers suffering from math anxiety or for the mathematical novice. For example, in the first 75 pages (about a quarter of the book) this author presents little more than how to form stem-and-leaf plots, boxplots, and a few descriptive statistics. More than a page is devoted to explain things as simple as rounding error, why the percentages in tables do not always add up to exactly 100%, and that "there is a big difference between 0.45 and 0.45%." Now either (1) the students I have taught tend to be much brighter, (2) I cover too much higher-level material in my classes (a claim that occasionally occurs in student reviews), or (3) Statistics for Archaeologists could have included more.
Admittedly, Drennan indicates in his preface that one group this book is aimed at includes those who find quantitative reasoning "difficult or intimidating." Additionally, there is somewhat of a literary tradition in archaeological statistics books that aims at the lowest common denominator in student skills. Clive Orton's Mathematics in Archaeology (William Collins Sons, London, 1980) barely presents more than a single equation in the entire book; Mike Fletcher and Galy R. Lock's Digging Numbers: Elementary Statistics for Archaeologists (Oxford University Committee for Archaeology, Oxford, 1991) covers more ground in greater mathematical detail than Orton, but falls somewhere below the level of presentation given by Drennan. Is there something wrong here? Must we instructors assume mathematical incompetence? While it is true that some percentage of students suffer from a mathematical phobia and do not have a clue as to why we might treat data quantitatively, I have found that catering to this level causes boredom among the majority of students. And, because so little material at such a low level is covered, an appreciation of what statistics can offer is never gained. In this practice we actually lose the upper end of student talent, and then wonder at the lack of mathematical sophistication in the discipline at large when the remaining students eventually become professionals.
It is paradoxical, then, that Statistics for Archaeologists rapidly grabbed my attention with an attack on mathematical ineptness and its social acceptability:
...an otherwise well-educated person can profess a complete inability to comprehend anything about numbers beyond addition and subtraction without incurring the disdain to be expected if he or she admitted to verbal skills so limited as to make everything in the daily newspaper but the comics unintelligible (p. viii).
Drennan sees mathematics not as an arcane side-line but as essential to doing archaeology; he emphasizes that archaeological data are "naturally numerical" and that as a consequence analysis is "unavoidably quantitative." What does the book cover? Much of it comes from John W. Tukey's EDA (Exploratory Data Analysis) material, particularly in the use of certain forms of graphical displays (stem-and-leaf plots, boxplots) and descriptive statistics (trimmed means and standard deviations, midspreads). Most of the book, however, overviews the "classical" statistical methods developed during the 1920s-1950s: one- and twosample t-tests, one-way ANOVA (analysis of variance), tests for tabulated data based on the chi-square, pearsonian correlation, regression, and the like. …