Comparison of three alternative methods for the analysis of interrupted time-series

This paper examines three procedures for analyzing interrupted time-series designs. Particularly, the functioning of correctly identified model, the g eneral transformation approach, and an alternative approach proposed by Esc udero and Vallejo (1999). Simulated data were used for investigating the ef fects that two intervention models had on the estimation of parameters, err or rates and power rates of 20 ARIMA processes. The results indicate that, under manipulated conditions: 1) the accuracy of estimates is essentially t he same for all three approaches. 2) The alternative approach controls the Type I error rate when alpha=0.05 and the model has more than one parameter : under the remaining conditions the model provides conservative rates of e rror Finally, with an immediate and constant intervention effect the altern ative approach is less powerful than the rest of the approaches: however, w ith an gradual and permanent intervention effect all three analyses provide d equivalent results.