Chapter 8 - Statistical Significance, Explanation, and Prediction

Abstract

Statistical significance does not necessarily imply clinical significance or clinical meaningfulness. Formal statistical testing indicates whether an observed outcome is likely to reflect the truth in the population and yet it remains silent on the importance of the phenomenon itself. Simple descriptors provide summary of the numbers and thus a reasonable starting point for getting a sense of the data. The answers provided by inferential statistics are always of the same form and constitute “statistical explanation.” The goal of inferential statistics is to understand phenomena by relating variation in them to variation in other phenomena. The stronger the relationship, the better the explanation, and whether the relationship is strong or weak, it is almost always incomplete. As such, prediction of parameters such as response to treatment cannot be perfect. Inferential statistics provides techniques that help explain phenomenon. Statistical explanation or statistical inference is done by relating variables to one another, specifically, by showing that variation in one variable is associated with variation in another. Inferential statistics provides techniques that enable determining whether or not the differences observed between two or more samples reflect the truth in the population or are likely due to chance. Statistics make use of numerous models, one of the more common of which is the normal distribution, which describes how some types of phenomenon behave.