Definition
A time-series dataset is a sequence of observations indexed by time, where both the order of the observations and the spacing between them carry information. Reordering a cross-sectional dataset loses nothing; reordering a time series destroys its meaning. This single property — temporal dependence — is what makes the shape distinctive and is the source of every technique that distinguishes time-series analysis from ordinary statistics.
The shape dominates whole disciplines. Asset prices, climate readings, server metrics, sensor telemetry, biomedical signals, and economic indicators all arrive as time-stamped observations. Whether the spacing is regular (one observation per second) or irregular (every quote update) shapes which analytical tools apply.
Why it matters
How it works
Working with a time series begins with three questions. Is the series stationary — do its statistical properties (mean, variance, autocorrelation) hold steady over time? If not, can a transformation such as differencing, log-returns, or detrending make it stationary so standard tools apply? What are the underlying components — is there a long-run trend, a recurring seasonal cycle, and a residual noise term? These questions are not academic; they determine which forecasting model is even applicable. ARIMA, exponential smoothing, state-space models, and neural sequence models each assume specific properties of the input.
The biggest practical hazard is look-ahead bias. Many natural-feeling analyses — normalising a series by its full-sample mean, fitting a model on all data and reporting in-sample fit, joining a daily price series to a fundamental metric whose true publication date lagged — silently use information from the future. The discipline that prevents this is point-in-time data handling: every value is tagged with the moment it would have been observable, and every calculation at time t may use only values whose observation time is at or before t. Walk-forward analysis and rolling-window evaluation are the operational tools that make this discipline enforceable.