Time series

A time series is a sequence of observations ordered by time, often recorded at regular intervals. Its temporal ordering makes trends, cycles, and dependence between observations central to its analysis.

Autocorrelation: The correlation between values of a variable and values of the same variable at different time lags. It measures how strongly a series depends on its own past.
Autoregressive model: A model that predicts a variable from its own previous values. It represents persistence directly through lagged observations.
Time-series forecasting: The use of past observations and temporal structure to estimate future values. Forecasting is a direct use of observed sequences for planning.
Cross-sectional data: Data collected across multiple entities at a single point or period in time. Unlike a time series, its central structure is variation among entities, not temporal order.
Stationary process: A stochastic process whose statistical properties remain unchanged under shifts in time. Stationarity determines whether stable patterns can be modeled across a series.
Moving-average model (time series): A time-series model that represents observations using current and past random shocks. It captures short-lived effects that echo through subsequent observations.
Anomaly detection: The identification of observations that differ substantially from expected patterns. Unexpected deviations can signal faults, attacks, or unusual events in a series.
Panel data: Data tracking multiple entities across multiple time periods. It combines temporal sequences with repeated measurements across entities.
Trend (time series): A long-term change in the average level of a time series. Separating trend from short-term variation clarifies the series’ underlying direction.
ARIMA: A forecasting model combining autoregression, differencing, and moving-average terms. It handles serial dependence after differencing removes certain trends.

Linked from 23 pages