Identifiability Limits in Density-Dependent Leslie Models Under Partial Observation: Linking demographic stochasticity to inference bias in structured population time series (Δt = 1)
Keywords:
Leslie matrix, population dynamics, density dependence, identifiability, state-space models, demographic stochasticity, JATS 1.1, deep section nesting, table footnotes, appendixAbstract
BACKGROUND: Structured population models are frequently inferred from incomplete demographic time series, yet parameter identifiability can fail even when models fit well. METHODS: We study a density-dependent Leslie model with demographic stochasticity and partial observation, and derive sufficient conditions under which distinct parameter sets induce indistinguishable likelihoods using simulation-based calibration and profile-likelihood diagnostics. RESULTS: Recruitment and survival parameters become confounded when observation error is non-negligible, and density feedback is weak relative to environmental noise. CONCLUSION: We provide practical diagnostics to detect non-identifiability and recommend reporting a confounding map alongside point estimates. This XML additionally functions as a JATS 1.1 stress-test fixture.
