Joint coverage probability in a simulation study on Continuous-Time Markov Chain parameter estimation.

Area of Science:

• Statistics
• Biostatistics
• Computational Biology

Background:

• Parameter dependency is common in simulation studies, particularly for Continuous-Time Markov Chains (CTMC).
• Existing literature lacks a thorough examination of estimation performance for general multi-state CTMC models, especially concerning parameter estimate dependencies.
• Previous studies have not accounted for the dependency among parameter estimates in their assessments.

Purpose of the Study:

• To develop a multivariate approach for evaluating accuracy and precision in simulation studies.
• To provide a comprehensive examination of the estimation performance of a general 3-state CTMC model.
• To address the gap in the literature regarding parameter dependency in CTMC estimation.

Main Methods:

• Conducted simulation studies analyzing longitudinal data with a trinomial outcome using a CTMC, with and without covariates.
• Calculated performance measures including bias, component-wise coverage probabilities, and joint coverage probabilities.
• Applied the methods to data on Alzheimer's disease caregiver stress levels.

Main Results:

• Comparisons between joint and component-wise parameter estimates produced conflicting inferential results in simulations.
• The presence or absence of covariates influenced the estimation performance and the observed conflicts.
• The developed multivariate approach offers a more nuanced assessment of accuracy and precision.

Conclusions:

• Caution is advised when conducting simulation studies to assess performance due to potential parameter dependencies.
• The choice of statistical inference method should align with the specific goals of the simulation study.
• Accurate assessment of CTMC models requires accounting for parameter estimate dependencies.