Change in the structure of a data-generating process, e.g., level shifts in a time series.
Structural change usually refers to time-series data. The most common form is a level shift, but changes in slope or variance are also often seen.
The challenge includes:
- Detecting historical structural change (see, e.g., the
strucchange
package for R and references therein for an introduction) - Treating historical structural change (e.g., using dummy variables)
- forecasting when the next structural change will occur, and of what magnitude