What we learn when our data are abnormal

Abstract Most biological data do not fit a bell curve (i.e., is non-Gaussian). Simply by looking at the specific shape of graphed data, it is often possible to draw profound conclusions about the meaning of the data and the relationships among variables. For example, multiple peaks in a data distribution always demand explanation. In many cases the multiple peaks indicate that several different classes of data object have been mixed together. Sometimes, multiple peaks indicate that the data were poorly collected. In this chapter, we will show that non-Gaussian data distributions often lead us to important biomedical discoveries. Examples will include establishing nonlinear relationships among variables, the differences in the types of diseases that occur in children compared with the diseases occurring in adults; young and old age of occurrence peaks in Hodgkin lymphoma. We will also see how we can remove the “outlier” status of a particular type of ovarian cancer, and shed some light on the origins of serous carcinoma of the ovary.