Statistical methods for risk-outcome research: being sensitive to longitudinal structure

Annu Rev Clin Psychol. 2009:5:71-96. doi: 10.1146/annurev-clinpsy-060508-130357.

Abstract

The relation between risk and outcome consists of myriad, complex, longitudinal processes. To study these relations requires research designs and statistical methods that are sensitive to the longitudinal structure of the risk, the outcome, and the risk-outcome relation. This review presents four longitudinal characteristics that can complicate psychopathology risk-outcome research. We represent each complication with an example data set. We demonstrate how conventional statistical approaches can yield highly misleading results. Finally, we review alternative statistical approaches that can handle these complications quite well.

MeSH terms

  • Adult
  • Humans
  • Logistic Models
  • Longitudinal Studies
  • Mathematical Computing
  • Mental Disorders / diagnosis
  • Mental Disorders / psychology*
  • Models, Statistical*
  • Risk Factors
  • Young Adult