MIAmaxent

MIAmaxent implements an ecologically interpretable framework for species distribution modeling by using subset selection and ecologically grounded variable transformations to produce parsimonious maximum entropy and logistic regression models.

Key Features:

• Subset selection over lasso regularization: Uses subset selection instead of lasso regularization (as in Maxent) to produce simpler, more parsimonious models.
• Variable transformation based on ecological theory: Transforms predictor variables according to expected occurrence–environment relationships to align models with ecological principles.
• Flexible model fitting approaches: Supports maximum entropy fitting for presence-only data and logistic regression (GLM) for presence–absence data.
• Decoupling of methodological steps: Separates variable transformation, model fitting, and model selection to enable comparative evaluation of approaches.
• Tools for data exploration and model interrogation: Provides functions to explore input data and interrogate model outputs in light of existing ecological knowledge.

Scientific Applications:

• Maximum entropy distribution modeling: Predicts species distributions from presence-only data with an emphasis on ecological interpretability.
• Presence–absence modeling with GLM: Models species distributions from presence–absence data using logistic regression (GLM).
• Biodiversity research and conservation planning: Supports habitat suitability analysis, biodiversity assessment, and conservation planning where understanding ecological drivers is important.

Methodology:

Implements subset selection and ecologically informed variable transformations, supports maximum entropy fitting for presence-only data and logistic regression (GLM) for presence–absence data, and decouples variable transformation, model fitting, and model selection while providing functions for data exploration and model interrogation.