Correlation and Regression Analysis: A Historian’s Guide by Thomas J. Archdeacon (review)

370 The Canadian Historical Review strange schema produces an often disconnected text of 237 pages containing no fewer than sixteen chapters, plus a prologue and epilogue. It produces, too1 notes that refuse to quit. Sometimes these furnish information that might better have been placed in the text. Other times they provide an avalanche of trivia. While the Godfreys' volume is generally free of 'typos,' at least one of their lengthy, cumbersome notes (289, n. 19) clearly defeated their attempt to proofread for logic and consistency. In it the two authors appear variously as 'I' and 'we.' While Search Out the Land is flawed, its close examination of the 'growth of equality' for the Jews of Britain's American colonies, 17401867 , is unlikely to be significantly revised, or even soon revisited. COLIN READ Huron College, University ofWestern Ontario Correlation and Regression Analysis: A Historian's Guide. THOMAS J. ARCHDEACON. Madison: University of Wisconsin Press 1995. Pp. xxi, 352. $50.00 cloth, $22.50 paper Nearly twenty-five years after the coming of the New Social History, historians in this country and elsewhere have by and large resisted the methodological attractions of social science methods. Of the fourteen articles in volume 76 of the Canadian Historical Review, for instance, only three contain tables, all of which simply present data. None is subjected to any explicit statistical test. The CHR ranks with the more conservative Canadian history publications in this regard, but even journals more replete with numerical data seldom include statistical tests of an explicit hypothesis, beyond the occasional Pearson correlation or chi-square. The resort to statistical techniques for testing models or for generalizing from samples to populations, methods that have gained wide acceptance in other social science disciplines, remains extremely rare among historians. Part of the reason, Wisconsin historian Thomas J. Archdeacon argues, is the 'shortage of materials designed to explain quantitative analysis to historians.' He offers this volume as a means to ameliorate the situation. The book is divided into three sections, progressively more technical and arduous. The first section proceeds on the assumption that the reader has no familiarity with statistics. It covers the definition of variables, the concepts of central tendency and distribution , and the different types of sampling. The second part explains correlation, describes measures of association for nominal variables, and provides a clear mathematical description of linear regression statistics and their substantive import. Book Reviews 371 Here the reader had better hone his ability to read equations, because the third part quickly becomes heavy reading. It covers methods for induding nominal independent variables in a regression equation, examines ways to transform data that are not linearly distributed into the linear distributions required for regression analysis, and shows how to test causal models with correlations. The final two chapters discuss logistic regression, used for analysing the influence of independent variables on dichotomous dependent variables, and loglinear analysis for dealing with variables that are all measured on a nominal scale. The results produced by the latter two techniques are not as intuitive as those of ordinary least squares regression, and the last two chapters prove to be the most demanding of the book. The learning curve is likely to be too steep for historians who are unacquainted with statistical methods. Most of the text is given over to the mathematical understanding of the various techniques presented in the book. Readers will find little historical insight to shore up their interest along the way. Examples are drawn mainly from economic history, and seem chosen more for their use oftechnique than for their substantive findings. Indeed, the author spends little time on the substance of historical arguments and on the relevance of the statistical methods adduced in their behalf. This is the major failing of the book. Statistical techniques serve to refute or to sustain hypotheses, and hypotheses are derived from explicit models and theory. While this methodology is acknowledged briefly in a few paragraphs dispersed in different chapters, it does not inform the presentation of statistical techniques, so the reader is left to his own devices in gauging the usefulness of the techniques to historical argumentation. Rarely are the hypotheses underlying the examples discussed explicitly for their historical...