Cox regression analysis
a statistical technique used to build multivariate models that relate one or more continuous or categorical variables to survival times, without requiring researchers to specify in advance the form or nature of such relationships. For example, one might use Cox regression to determine how likely it is that alcoholics who are abstinent at 3 months and at 6 months will relapse. A key methodological concept in Cox regression analysis is the hazard, that is, the immediate potential or “risk,” of event occurrence. There are two types of Cox regression: the simpler standard Cox regression model (or proportional hazards model) and a more complex generalization known as the extended Cox regression model. The standard model is used when the risk of event occurrence for the reference and comparison groups remains constant relative to one another over all time points, whereas the extended model is used when the effect of particular variables
upon the occurrence of the event of interest changes over time. [David R. Cox (1924– ), British statistician]