Categorical Data

Abstract

A major objective of clinical research is to improve the effectiveness of individual therapies by studying treatment effects and health effects in subgroups of patients, for example age-groups, races, genders etc. Sometimes, the use of continuous or binary variables are possible for the purpose. However, races, numbers of co-medications, co-morbidities and many more variables in clinical research have stepping functions with a limited number of values, e.g., four races, zero to eight co-medications etc. If such stepping functions are analyzed using continuous variables in a linear or logistic regression model, we assume that the outcome variable will rise linearly, but this needs not necessarily be so. This assumption raises the risk of underestimating the effects. In the given situation, it may be more safe to recode the stepping variables into the form of categorical variables.