Gerhard Tintner’s Review of Methods of Correlation Analysis, Second Edition (1942)

Gerhard Tintner’s book review assesses the second edition of Mordecai Ezekiel’s Methods of Correlation Analysis as an influential practical text whose revisions strengthen its usefulness without fundamentally changing its approach. The review moves from the book’s established importance to its new chapters, its treatment of statistical reliability, and a qualified endorsement. Its central distinction is between improvements in the exposition and application of correlation methods and theoretical problems that remain unresolved within statistics itself.

Tintner begins by locating Ezekiel’s authority in empirical research:

There is hardly a book which has exerted so much influence on empirical studies by economic statisticians and in related fields using the techniques so lucidly presented in this book.

The survey of applications in chapter xxiii serves as evidence of that influence: Tintner regards the studies listed there as substantially indebted to Ezekiel’s methods. Yet influence does not imply methodological transformation. The second edition retains the earlier approach, adding somewhat greater emphasis on distinguishing a population from a sample. Tintner thus presents the revision as a consolidation of an established research instrument.

The two new chapters extend its practical reach in different directions. Chapter xix treats the reliability of individual forecasts in time-series analysis, presenting Henry Schultz’s pioneering work on the standard error of a forecast from a curve and applying it to simple and multiple correlation. Tintner finds the discussion of error formulas sketchy and disappointing, but carefully separates that limitation from the author’s responsibility:

The author, however, is not to blame for the fact that there is as yet no generally accepted solution to many of the problems connected with this very difficult question.

This qualification makes the review more than an inventory of additions. It identifies a boundary between usable procedures and unsettled foundations: a textbook cannot supply an agreed solution where the discipline has none. By contrast, chapter xvi offers computational and graphical expedients, presenting L. H. Bean’s short-cut methods for determining net regression lines and curves, with applications to simple and curvilinear regression and an extensive bibliography on graphic methods.

Tintner’s other reservation concerns statistical inference. Chapters ii and xviii expand their discussions of the reliability of statistical results and correlation conclusions. Nevertheless, they chiefly present R. A. Fisher’s position on fiducial limits, leaving aside newer developments in estimation and hypothesis testing associated with Neymann, Egon Pearson, and Wald. Tintner’s parenthetical suggestion that this omission may be wise tempers the criticism without erasing it.

The final judgment remains affirmative:

The changes made represent in general genuine improvements.

Tintner expects the edition to increase Ezekiel’s influence in applied statistics, while hoping a third edition can incorporate solutions to currently unresolved problems. The review’s significance lies in this balanced appraisal of practical accessibility, methodological continuity, and incomplete theoretical coverage.