This book is intended for those involved in data analysis in diverse areas of research. Unlike in a well-controlled and well-designed statistical experiment, many of us face data to which the notion of “data being a random sample from the normal population” does not apply. “Normal distribution” means that data must be continuous, but most data we deal with in the social sciences are categorical or non-numerical. Furthermore, we know that the relation between two normally distributed variables is by design linear. In practice, however, we encounter many nonlinear relations such as “the strength of the body is generally a concave function of age.” Such a phenomenon exists and cannot be ignored for the sake of using the normal distribution assumption.