Spatial Interrupted Time Series
Spatial Interrupted Time Series Analysis · Also known as: Spatial ITS, Geospatial ITS, Spatially-adjusted ITS, SITS
Spatial Interrupted Time Series (Spatial ITS) extends the classic ITS design to settings where units are geo-referenced and outcomes in one location may spill over into or correlate with outcomes in neighbouring locations. It estimates the causal effect of a discrete intervention on an outcome time series while explicitly modelling geographic autocorrelation, preventing biased standard errors and enabling detection of spatial spillovers.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use Spatial ITS when you have a clearly defined discrete intervention affecting geo-referenced units observed over time (typically 12 or more pre-intervention time points per unit), and when Moran's I on residuals or theory suggests spatial autocorrelation. It is well suited to public health surveillance data, environmental policy evaluation, and regional economic policy analysis. Do not use it when units are truly spatially independent (a spatial model adds unnecessary complexity), when the number of pre-interruption time points is fewer than 8–10, or when intervention timing varies widely across units without a common breakpoint (prefer staggered DiD instead).
Strengths & limitations
- Provides unbiased standard errors when geographic autocorrelation is present, avoiding the inflated significance that naive ITS produces.
- Explicitly quantifies and tests spatial spillover effects, which are often substantively important in policy evaluation.
- Retains the interpretability of the classic ITS design — level change and slope change remain the primary estimates.
- Applicable to aggregate geo-referenced data (regions, districts, hospitals) without individual-level records.
- Can accommodate spatial panel structures, allowing unit and time fixed effects alongside the spatial correction.
- Requires a sufficient pre-intervention time series (ideally 12+ points) for reliable trend estimation; short series produce unstable slope estimates.
- The choice of spatial weights matrix W is subjective; results can be sensitive to the weighting scheme, and no single choice is universally correct.
- Assumes a single common breakpoint; policies phased in at different times across units violate this and require staggered-adoption extensions.
- Software support is narrower than for standard ITS, and combined spatial-ITS estimation demands familiarity with spatial econometrics packages.
- Spillover testing is confirmatory rather than causal — detecting a change in neighbours does not isolate the mechanism of spillover.
Frequently asked
When should I use Spatial ITS instead of standard ITS?
Use Spatial ITS whenever your units are geo-referenced and a Moran's I test on ITS residuals suggests spatial autocorrelation (p < 0.05). If units are truly independent — for example, widely separated clinics with no patient overlap — standard ITS is sufficient and simpler.
How many pre-intervention time points do I need?
At least 8–10, preferably 12 or more, to estimate a stable baseline trend. Fewer points make the slope parameter imprecise and the level-change estimate unreliable, regardless of the spatial correction.
What if the intervention was rolled out at different times in different areas?
A common breakpoint is assumed. Staggered rollout violates this and requires staggered-adoption difference-in-differences or a panel ITS model with unit-specific breakpoints rather than Spatial ITS in its standard form.
How do I choose the spatial weights matrix?
Base the choice on theory or data structure before seeing results. Contiguity (shared borders) is natural for administrative regions; distance decay is appropriate when the mechanism operates over a gradient. Sensitivity analyses comparing two or three plausible matrices are good practice.
What does the spatial autocorrelation parameter rho mean?
Rho measures the degree to which outcomes in one unit co-move with the weighted average of neighbours after controlling for the ITS trend and intervention. A large positive rho confirms strong spatial clustering; a rho near zero means the spatial correction had little effect.
Sources
- McDowall, D., McCleary, R., Meidinger, E. E., & Hay, R. A. (1980). Interrupted Time Series Analysis. Sage Publications. ISBN: 978-0803913950
- Lawson, A. B. (2018). Bayesian Disease Mapping: Hierarchical Modeling in Spatial Epidemiology (3rd ed.). CRC Press. ISBN: 978-1138575424
How to cite this page
ScholarGate. (2026, June 3). Spatial Interrupted Time Series Analysis. ScholarGate. https://scholargate.app/en/causal-inference/spatial-interrupted-time-series
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Interrupted Time SeriesCausal inference↔ compare
- Panel Data Interrupted Time SeriesCausal inference↔ compare
- Spatial Causal Impact AnalysisCausal inference↔ compare
- Spatial Difference-in-DifferencesCausal inference↔ compare
- Spatial Propensity Score MatchingCausal inference↔ compare
- Spatial Regression Discontinuity DesignCausal inference↔ compare