An Efficient Implementation of the EMICM Algorithm for the Interval Censored NPMLE

ABSTRACT

The EMICM algorithm is an established method for computing the interval-censored NPMLE, a generalization of the Kaplan Meier curves for interval censored data. The novel contribution in this work is an efficient implementation, allowing each iteration to be computed in linear time. Using simulated data, it is shown that this new implementation is significantly faster than alternative EMICM implementations or other competing algorithms, allowing for analyses of datasets orders of magnitude larger than previously available.

KEYWORDS:

• EMICM
• Interval-censoring
• NPMLE
• Turnbull estimator