Obtaining Interval Estimates of Policy Impacts From Interrupted Time Series
Abstract
The absence of a means for obtaining confidence intervals for the impact of an intervention represents a serious gap in the literature on the analysis of interrupted time-series designs. In this article, an approximate procedure for constructing such confidence intervals is developed both in general terms and for two specific forms of intervention effects—abrupt temporary effects and gradual permanent effects. The adequacy of the procedure is examined using Monte Carlo simulation. An illustration of the procedure using a real data set is presented.