Simple statistical models can be sufficient for testing hypotheses with population time-series data

Area of Science:

• Ecology
• Population Dynamics
• Statistical Modeling

Background:

• Time-series data are crucial for understanding factors influencing species abundances.
• Sophisticated models often require unavailable species detectability data.
• Simpler models may be adequate for hypothesis testing in abundance time-series analysis.

Purpose of the Study:

• To evaluate the adequacy of simpler regression models for analyzing species abundance time-series data.
• To compare three regression models (A, B, C) using simulated and empirical datasets.
• To determine model suitability based on species life-history strategies.

Main Methods:

• Compared three regression models: conventional generalized linear model (A), autoregressive model (B), and population growth rate model (C).
• Utilized simulated and empirical datasets (fish and mammals).
• Employed both Bayesian and non-Bayesian fitting methods.

Main Results:

• Model C showed stronger support for K strategists (long-lived, low-fecundity), while Model A suited r strategists (short-lived, high-fecundity).
• All models (A, B, C) were supported for different species in real-world analyses, sometimes yielding distinct insights.
• Model C revealed predictor variable effects not apparent in Models A and B; Bayesian and frequentist approaches yielded similar results.

Conclusions:

• Relatively simple models are valuable for hypothesis testing in abundance time-series data, especially when complex model data is lacking.
• Fitting multiple models can provide complementary ecological insights.
• Model C, focusing on population growth rate, can uncover effects missed by simpler abundance models.