| The history of multivariate statistics, considered as a distinct subject, has been given relatively little attention by historians of statistics. Although details of the history of multivariate analyses are habitually included in more extensive treatises on the general history of statistics (e.g., Desrosieres, 1998, Porter, 1986; Stigler, 1986), the story of the social, political, and methodological incentives that gave rise to multivariate methodology in particular, has as yet, gone untold.; The present work attempts to fill this void by providing an in-depth historical analysis of the social, political, and methodological context in which multivariate methodology arose, beginning in late nineteenth century Britain. The discovery of correlation by Francis Galton in the 1880s gave birth to a revolution in statistical work, as scientists and statisticians of late nineteenth century grappled with how to use and extend Galton's bivariate method. George Udny Yule, in 1899, would give what is usually considered to be the first complete econometric analysis in the history of statistics (Desrosieres, 1998). As will be seen, the advent of both multiple regression in late nineteenth century and Sewall Wright's path analysis in the early 1920s were both inspired more by applied research problems than they were by drives to extend statistical methodology for its own sake.; An analysis of the historical context in which multivariate methods arose should be of interest to both the historian of statistics and to the applied statistician, for it provides a knowledge and appreciation of statistics that cannot be otherwise obtained. |
