Practical advice on variable selection and reporting using Akaike information criterion.

Area of Science:

• Ecology
• Statistics
• Ecological Modeling

Background:

• Model selection is crucial in ecological research, with the Akaike information criterion (AIC) being a dominant tool.
• Common misunderstandings persist among users regarding AIC application, interpretation, and reporting.
• Specific confusion exists around 'pretending' variables and the role of p-values in AIC-based model selection.

Purpose of the Study:

• To address prevalent user misconceptions surrounding the Akaike information criterion (AIC).
• To provide intuitive understanding of AIC application and interpretation through simulation.
• To promote improved statistical practices in ecological model selection and reporting.

Main Methods:

• The study complements existing technical literature on AIC.
• Simulation methods are employed to develop intuition around AIC concepts.
• Focus is placed on interpreting AIC model tables and the relationship between p-values and AIC.

Main Results:

• Simulations provide practical insights into the application of AIC.
• Clarification is offered on the concept of 'pretending' variables in model selection.
• Guidance is provided on the interpretation of statistical support when using AIC.

Conclusions:

• Enhanced understanding of AIC can lead to more robust ecological modeling.
• Simulation-based intuition aids in overcoming common statistical pitfalls.
• The study advocates for better practices in using, interpreting, and reporting AIC-selected models.