Profile likelihood-based parameter and predictive interval analysis guides model choice for ecological population...

Area of Science:

• Ecological modeling
• Mathematical biology
• Coral reef ecology

Background:

• Mathematical models are crucial for ecological data analysis, but face a trade-off between data availability and model complexity.
• Standardized data collection is often lacking in ecological studies, complicating model selection.
• Choosing appropriate ecological models typically requires case-by-case evaluation.

Purpose of the Study:

• To quantitatively explore the trade-off between data availability and model complexity in ecological modeling.
• To compare the predictive performance of a simple single-species ordinary differential equation (ODE) model with a more complex two-species coupled ODE model for coral reef regrowth.
• To introduce and validate a new parameter-wise prediction approximation method for assessing model predictability.

Main Methods:

• Utilized univariate profile likelihood analysis to assess model identifiability.
• Constructed and compared approximate prediction intervals using a novel parameter-wise prediction approximation.
• Validated the approximation by comparing it against a rigorous evaluation of the full likelihood.

Main Results:

• Both the simple and complex ordinary differential equation (ODE) models were found to be practically identifiable.
• The approximate parameter-wise prediction interval analysis revealed similar predictive performance between the simple and complex models.
• The new approximation method provided explicit insights into how each parameter influences model predictions.

Conclusions:

• The developed parameter-wise prediction approximation is a reasonable and computationally efficient method for evaluating ecological model predictions.
• The study demonstrates that simpler models can be as effective as complex ones for prediction in certain ecological scenarios.
• Freely available algorithms and software in Jupyter notebooks facilitate adaptation for other ordinary differential equation (ODE)-based ecological models.