Time-series analysis

A time series is a sequence of observations ordered through time. Biological examples include daily case counts, population abundance, hormone concentrations and environmental measurements.

Dependence through time

Successive observations are often correlated, so methods assuming independent observations may be inappropriate.

The lag-\(k\) autocorrelation measures association between observations separated by \(k\) time steps.

Components

Time series may contain trends, seasonal patterns, cycles, interventions and irregular fluctuations. These components should be distinguished from mechanistic biological dynamics where possible.

Forecasting

A fitted time-series model can generate forecasts with uncertainty intervals, but predictive performance should be assessed using later or held-out observations.