Minimum Message Length Inference of the Exponential Distribution with Type I Censoring

Area of Science:

• Statistics
• Biostatistics
• Data Science

Background:

• Censored data is prevalent across scientific disciplines.
• Statistical models are typically estimated using maximum likelihood and information criteria like Akaike's Information Criterion.
• Existing methods face challenges with specific types of censoring.

Purpose of the Study:

• To demonstrate the application of the information theoretic minimum message length (MML) principle for estimating statistical models.
• To address challenges posed by type I random and fixed censoring in data analysis.
• To compare MML estimates with standard maximum likelihood estimates, particularly for mean survival time.

Main Methods:

• Utilized the minimum message length (MML) principle for statistical model estimation.
• Applied the MML principle to data with type I random censoring.
• Applied the MML principle to data with type I fixed censoring.
• Employed the exponential distribution as a case study for demonstrating the methodology.

Main Results:

• Successfully estimated statistical models using the MML principle under both random and fixed censoring conditions.
• The exponential distribution with censoring was effectively modeled using MML.
• Minimum message length estimation of mean survival time demonstrated advantages over maximum likelihood estimation.

Conclusions:

• The minimum message length principle provides a viable and advantageous alternative for estimating statistical models with censored data.
• MML offers improved estimation of mean survival time compared to maximum likelihood in the presence of censoring.
• This approach enhances statistical modeling capabilities in fields dealing with censored data.