The Biostatistics Journal Club will meet on Tuesday, November 3rd, 2015
COPH 2280
12:00 – 1:00pm.
James P. Selig, Ph.D..
“Beyond a World of Straight Lines: Thoughts, Tools, and Tips for Using Statistical Models to Study Nonlinear Change.”
Abstract
One of the many exciting things we can do with longitudinal data is study how things change over time. Change is central to research questions across many fields. For example, we might want to know more about: change in the number of cancer cells; how quickly a drug is metabolized; how vocabulary increases for young children, or how quickly we forget things.
Often researchers examine such changes using models based on linear (straight line) or non-linear mathematical functions. These models can be applied to describe both how variables change over time and what other variables are predictors or sequelae of change over time.
Despite the fact that many phenomena are best described by nonlinear functions, it is common practice in some fields to almost always use models for linear change. This may be due to the fact that models for nonlinear change are less well known or because the complexity of models for nonlinear change can be overwhelming.
This presentation will introduce the application of models for nonlinear change. I will emphasize using the simplest possible models and choosing models that best address key research questions. I will cover the following topics: a) distinctions between linear and nonlinear change and between linear and nonlinear models; b) desirable features for models of nonlinear change; c) using ‘linear’ models to study nonlinear change, and d) tools for estimating a variety of additional parameters to better understand change.