Tobit regression for modeling mean survival time using data subject to multiple sources of censoring

We propose tobit regression methods based on weighted maximum likelihood which are applicable to survival times subject to both fixed and random censoring times

Qi Gong; Douglas E. Schaubel

2018

Scholarcy highlights

  • In biomedical studies, Cox regression is the most popular modeling approach for the analysis of censored data in the presence of covariate adjustment
  • It is possible that the investigator may prefer to describe the results of a survival analysis to a stakeholder in terms of differences in mean lifetime
  • The Cox model is inappropriate in a lot of settings due to violation of the proportional hazards assumption
  • In the presence of proportionality, each Cox model parameter pertains to an ordering of survival functions, which is a very useful property
  • We essentially propose a weighted version of a complete-case analysis, fitting the Tobit regression model to only subjects either observed to die or observed to live until their fixed end-offollow-up time
  • We propose carrying out an inverseweighted complete-case analysis of {i : ∆Di + ∆Li = 1}; i.e., a complete-case analysis of the data that would suffice in the absence of C, inverse weighted such that the weighted data set represents the data that would be observed in the absence of C
  • This would often be much more convenient than what is currently a frequently applied alternative: modeling the hazard, combining the regression parameter and cumulative baseline hazard estimates, transforming and integrating the subject-specific survival curve

Need more features? Save interactive summary cards to your Scholarcy Library.