Composite likelihood inference for space-time point processes.

Area of Science:

• Ecology
• Statistical Modeling
• Forest Dynamics

Background:

• Rainforest ecosystems exhibit complex dynamics driven by tree births, deaths, and intricate interactions.
• Understanding these dynamics is crucial for conservation and predicting forest responses to environmental changes.
• Existing methods may struggle with the spatio-temporal complexity and data limitations inherent in rainforest censuses.

Purpose of the Study:

• To develop novel statistical regression models for analyzing rainforest tree recruitment and mortality patterns.
• To estimate model parameters using conditional composite likelihood functions, minimizing assumptions.
• To address challenges posed by short time-series and large spatial domains in rainforest data.

Main Methods:

• Specification of regression models for conditional intensity of recruits and probability of death.
• Estimation via conditional composite likelihood functions, focusing on first-order properties.
• Application of a central limit theorem in a fixed time-span, increasing spatial domain setting for asymptotic results.

Main Results:

• The proposed conditional composite likelihood method yields assumption-lean estimators for regression parameters.
• The methodology effectively handles stochastic covariates derived from past data.
• Weak dependence assumptions on the space-time process innovations are sufficient for valid inference.

Conclusions:

• The developed statistical framework provides a robust method for analyzing complex rainforest tree dynamics.
• This approach is suitable for datasets with limited temporal but extensive spatial information.
• The findings contribute to improved ecological modeling and understanding of forest regeneration and mortality.