ABSTRACT

There is much current research in the machine learning and statistics communities on algorithms for discovering knowledge and structure in data. Although many scholars (e.g., Selvin & Stuart, 1966) in the statistics community in the 1960s and 1970s considered such data-exploration activities as fishing or data dredging, Tukey (1977) argued that statistical theory needed to adapt to the scientific method. More than two decades hence, it appears that the statistics community has adopted Tukey’s perspectives and acknowledged that model search is a critical and unavoidable step in the model-fitting process (Glymour, Madigan, Pregibon, & Smyth, 1997).