Non-linear Modeling on a Pocket Calculator

Abstract

Non-linear relationships in clinical research are often linear after logarithmic transformations. Odds ratios, log likelihood ratios, Markov models and many regression models are models that make use of it. An example with real data is given. We have to add that logarithmic transformation is not always successful, and that alternative methods are available like Box Cox transformation, and computationally intensive methods like spline and Loess modeling. However, these methods generally require statistical software and cannot be executed on a pocket calculator. This chapter assesses simply logarithmic transformation of the outcome variable for linearization of survival data.