Chapter

21 Survival Analysis

Research Methods in Psychology

IV. DATA ANALYSIS ISSUES

  1. John B. Willett PhD1,
  2. Judith D. Singer PhD2,
  3. Suzanne E. Graham EdD3

Published Online: 26 SEP 2012

DOI: 10.1002/9781118133880.hop202021

Handbook of Psychology, Second Edition

Handbook of Psychology, Second Edition

How to Cite

Willett, J. B., Singer, J. D. and Graham, S. E. 2012. Survival Analysis. Handbook of Psychology, Second Edition. 2:IV:21.

Author Information

  1. 1

    Harvard University, Graduate School of Education, Cambridge, Massachusetts, USA

  2. 2

    Harvard University, Office of the Provost Holyoke Center, Cambridge, Massachusetts, USA

  3. 3

    University of New Hampshire, Department of Education, Durham, New Hampshire, USA

Publication History

  1. Published Online: 26 SEP 2012

Abstract

Psychologists often ask whether-and, if so, when-critical events occur, over the life-course. Such questions about event occurrence present unique design and analytic difficulties. The core problem is that no matter when data collection begins and no matter its length of follow-up, some participants may not experience the target event while observed. The event-histories of such participants are said to be censored. Although the prospect of censoring complicates research design and the presence of censoring complicates statistical analysis, the censored cases themselves reveal important insights into patterns of event occurrence. Why? Because these cases provide information on how long a subset of participants can continue without experiencing the event of interest. Consequently, they cannot be ignored but must be incorporated into data-analyses in some reasonable and informative way, if event histories are to be modeled unbiasedly. Typically, this is achieved by application of the methods of survival analysis. These methods permit the investigator to specify sensible statistical models for the risk of event occurrence over time and to compare these patterns among groups, while controlling for covariates, even in the face of censoring. Data collection can be prospective or retrospective; research designs can be experimental or observational. The passage of time can be measured continuously or discretely. In this chapter, we present a nonmathematical introduction to survival analysis. After describing the conceptual and statistical “building blocks” of the approach, we focus on important issues in research design and data analysis. In each case, we survey the problems that must be faced when adapting the methodology to psychological research and we provide guidelines for making informed decisions. In the process, we review how psychologists have used survival analysis successfully in the past and we point out new applications of the methods in the future. We conclude with recommendations for follow-up reading.

Keywords:

  • data collection;
  • research design;
  • survival analysis