Prediction accuracy measures for time-to-event models with left-truncated and right-censored data

ABSTRACT

Many time-to-event models have been developed for left-truncated and right-censored (LTRC) data, which arise in many applications involving follow-up studies. However, there is no work on evaluating the prediction accuracy of the time-to-event models for LTRC data. This paper develops two novel weighted prediction summary measures for a nonlinear prediction function with LTRC data. They are based on a weighted variance decomposition and a weighted prediction error decomposition, by the inverse probability weighting technique. The resulting measures are shown to be consistent and asymptotically normal. Simulation studies are conducted to evaluate their good finite sample performance. An empirical application to the Channing House data set illustrates the methodology.

Keywords:

• Prediction accuracy measures
• left-truncated and right-censored data
• inverse probability weighting technique
• variance decomposition
• predict error decomposition

Acknowledgments

Zhang's work was supported in part by the National Natural Science Foundation of China (11771133). Fan's work was partially supported by the National Natural Science Foundation of China (11801360), the Key Program in the National statistical science Research of China (2018LZ02).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

Zhang's work was supported in part by the National Natural Science Foundation of China [11771133]. Fan's work was partially supported by the National Natural Science Foundation of China [11801360], the Key Program in the National statistical science Research of China [2018LZ02].