Longer time series with missing data improve parameter estimation in a state-space model in coral reef fish...

Area of Science:

• Ecology
• Time Series Analysis
• Ecological Modeling

Background:

• Ecological time series analysis is crucial for understanding community dynamics.
• Missing data in ecological datasets are common due to various monitoring challenges.
• State-space models can handle missing data by estimating unobserved abundances.

Purpose of the Study:

• To evaluate the impact of missing data on parameter accuracy and precision in multivariate state-space models.
• To assess the performance of these models with varying time series lengths and sampling frequencies.
• To investigate the influence of spatial replication and sampling design on parameter estimation.

Main Methods:

• Utilized a multivariate process-based state-space model.
• Employed simulated data informed by Great Barrier Reef reef fish community counts.
• Compared model performance across different time series lengths, sampling frequencies, and spatial replication levels.

Main Results:

• Models fitted to longer time series with missing data yielded more accurate and precise parameter estimates than shorter, complete-case series.
• Higher spatial replication with uneven sampling intervals improved parameter precision.
• Analysis of reef fish data showed changes in population growth and density dependence parameters with longer, variable time series.

Conclusions:

• Bayesian state-space models demonstrate robustness to missing data in ecological time series.
• Flexibility in long-term monitoring sampling designs does not necessarily compromise ecological inference.
• These models offer reliable parameter estimates even with incomplete ecological datasets.