Fast estimation of regression parameters in a broken-stick model for longitudinal data

Area of Science:

• Statistics
• Biostatistics
• Computational Biology

Background:

• Change-point estimation in broken-stick models is vital for biological phenomena.
• Existing methods can be computationally intensive, particularly for multiple change-points.

Purpose of the Study:

• To develop a computationally economical, likelihood-based approach for estimating change-points in broken-stick models.
• To enhance efficiency in both cross-sectional and longitudinal data analysis.
• To provide a method for statistically sound inference.

Main Methods:

• A novel approach based on local smoothing in a shrinking neighborhood of each change-point.
• Likelihood-based estimation for efficient change-point detection.
• Simulations to compare computational viability against existing search-based methods.

Main Results:

• The proposed method demonstrates superior computational viability compared to existing techniques, with substantial gains in multiple change-point scenarios.
• Estimates exhibit [Formula: see text]-consistency and asymptotic normality.
• Asymptotic efficiency is achieved in the cross-sectional setting, facilitating robust statistical inference.

Conclusions:

• The new method offers a computationally efficient and statistically reliable tool for change-point estimation in broken-stick models.
• Applicable to both cross-sectional and longitudinal biological data, including hormone level changes and plant growth.
• Facilitates meaningful statistical inference for complex biological modeling.