Statistics should be an organic component of substantive investigation. This is how statistics should be learned—and how it should be taught.
A text should aim to give students what they will later need to know. What students will later need to know is how to utilize statistics in their empirical work. To get such transfer requires that statistics be embedded within a framework of substantive inquiry.
Substantive investigation rests on extrastatistical inference—substantive considerations concerning validity of task-procedure and generality of results. Practical understanding—transfer to empirical analysis—requires that statistics be integrated into a larger framework of extrastatistical inference.
This extrastatistical theme is embodied in the Experimental Pyramid of Figure 1.1. The six levels of the Pyramid portray a hierarchy of considerations involved in empirical investigation. Statistics, to be effective, needs to be integrated into the substantive considerations at each level of the Pyramid.
The main value of statistics is in planning the investigation, long before the data are collected. Contrary to the standard stereotype, the main function of statistics is to get more information into the data.
Current texts pursue two largely incompatible goals: To be a text for first-year graduate students and to serve as a reference handbook for advanced researchers. Both audiences suffer thereby, especially first-year students. Facing a plethora of formulas, uncertain which are basic, doing exercises largely devoted to numerical calculations, first-year students are hindered and side-tracked from developing understanding and research judgment.
Chapters 1–12 present a core intended for first-year graduate students. Far fewer formulas are presented than in other texts, which is intended to facilitate conceptual and empirical understanding. Two novelties are the heavy emphasis on confidence intervals and the separate chapters on confounding and single subject design.
Chapters 13–21 serve in part a reference handbook function, for they take up more specialized topics: within-versus-between design, Latin squares, multiple regression, analysis of covariance, quasi-experimental design, multiple comparisons, and the difficult problem of measuring effect size. Also included are chapters on the foundations of statistics, on mathematical models in psychology, and on psychological measurement theory. These chapters aim to give conceptual understanding that will facilitate empirical analysis.
The “Empirical Direction” in this book is not essentially new. It is a return to and unification of statistics with the extrastatistical nature of empirical science. Further discussion is given in Chapter 23, Lifelong Learning.