Interrupted Time Series Calculator

Analyze policy impacts with flexible interrupted series diagnostics. Enter observations, intervention timing, and optional labels. Get charts, coefficients, forecasts, and exportable reports instantly today.

Calculator Form

Paste outcome values, set the intervention point, and generate interrupted time series statistics in one stacked page layout.

Example: 7 means the seventh observation starts the intervention.
Adds future fitted values beyond the last observed point.
Controls number formatting in tables and summary cards.
Enter one value per line or separate values with commas.
Use commas or line breaks. Leave empty to auto-create observation labels.

Example Data Table

This sample shows a baseline trend before observation 7 and a stronger level after the intervention begins.

Label Time Outcome Intervention Active
Jan 1 102 No
Feb 2 104 No
Mar 3 103 No
Apr 4 105 No
May 5 106 No
Jun 6 108 No
Jul 7 115 Yes
Aug 8 117 Yes
Sep 9 119 Yes
Oct 10 121 Yes
Nov 11 122 Yes
Dec 12 124 Yes

Tip: press “Load Example Data” to populate the calculator instantly.

Formula Used

The calculator uses segmented linear regression, a standard interrupted time series approach:

Y_t = β0 + β1(Time_t) + β2(Intervention_t) + β3(PostTime_t) + ε_t
  • Y_t = outcome at time t
  • β0 = baseline intercept
  • β1 = pre-intervention slope
  • Intervention_t = 0 before intervention, 1 after it starts
  • β2 = immediate level change at intervention
  • PostTime_t = 0 before intervention, then counts time after intervention
  • β3 = change in slope after intervention

Counterfactual values are generated from the baseline trend only: β0 + β1(Time_t). Comparing fitted and counterfactual lines shows the estimated intervention impact.

How to Use This Calculator

  1. Enter your observed outcome values in time order.
  2. Set the observation number where the intervention begins.
  3. Optionally enter custom labels such as months or weeks.
  4. Choose how many future periods you want forecasted.
  5. Click the calculate button to estimate level and slope changes.
  6. Review the summary cards, coefficient table, and graph.
  7. Download the fitted dataset as CSV or PDF if needed.

Interpretation Notes

  • A positive Level Change suggests an immediate upward jump at intervention start.
  • A positive Slope Change suggests the post-intervention trend rises faster than before.
  • R-squared shows how much variation the segmented model explains.
  • Durbin-Watson near 2 usually suggests weaker autocorrelation concerns.
  • Consider seasonal patterns, delayed effects, and outside shocks when interpreting results.

FAQs

1) What is an interrupted time series design?

An interrupted time series design tracks repeated outcomes over time, then tests whether a defined intervention changed the level, trend, or both. It is useful when randomized experiments are not practical but a clear intervention date exists.

2) What is one difference between a time-series design and an interrupted time-series design?

A regular time-series design studies patterns across time without requiring a known intervention point. An interrupted time-series design explicitly includes a policy, event, or treatment date and estimates whether that interruption changed the outcome path.

3) How many observations should I collect?

More observations improve stability. A practical minimum is several points before and after intervention, but longer series are better. Very short series can produce unstable slopes and wide confidence intervals.

4) What does level change mean?

Level change is the immediate jump or drop when the intervention begins. It captures an abrupt shift in the outcome at the interruption point, separate from any later trend change.

5) What does slope change mean?

Slope change measures whether the post-intervention trend becomes steeper or flatter than the baseline trend. It tells you whether the intervention changed the rate of movement over time, not just the starting level.

6) Can seasonality affect results?

Yes. If your data follow monthly, weekly, or quarterly cycles, seasonality can bias estimates. In those cases, extend the model with seasonal indicators or additional structure before making strong causal interpretations.

7) When is this calculator not ideal?

This calculator is less suitable when you have strong autocorrelation, multiple overlapping interventions, missing time order, major seasonal effects, or non-linear relationships that a simple segmented linear model cannot capture well.

8) How should I choose the intervention point?

Choose the first observation where the intervention truly begins affecting the outcome. If implementation was delayed or phased, test alternative dates and compare results carefully rather than assuming a single exact breakpoint.

Related Calculators

Important Note: All the Calculators listed in this site are for educational purpose only and we do not guarentee the accuracy of results. Please do consult with other sources as well.