Tutorial

Survival Analysis—A Statistic for Clinical, Efficacy, and Theoretical Applications

gruberfa{at}hal.lamar.edu

Current demands for increased research attention to therapeutic efficacy, efficiency, and also for improved developmental models call for analysis of longitudinal outcome data. Statistical treatment of longitudinal speech and language data is difficult, but there is a family of statistical techniques in common use in medicine, actuarial science, manufacturing, and sociology that has not been used in speech or language research. Survival analysis is introduced as a method that avoids many of the statistical problems of other techniques because it treats time as the outcome. In survival analysis, probabilities are calculated not just for groups but also for individuals in a group. This is a major advantage for clinical work. This paper provides a basic introduction to nonparametric and semiparametric survival analysis using speech outcomes as examples. A brief discussion of potential conflicts between actuarial analysis and clinical intuition is also provided.

KEY WORDS: survival analysis, efficacy, prediction, phonology, language

Submitted on March 10, 1998
Accepted on July 23, 1998

CiteULike    Connotea    Del.icio.us    Digg    Facebook    Reddit    Technorati    Twitter    What's this?