Abstract
Evaluation of the effects of intervention in designs of time series in the presence of trends. The most typical solution to the problem of serial dependence has been the use of the ARIMA models proposed by Box and Tiao (1965). But, for a reliable use of those models, too many observations are required, witch implies that they can't be properly used in many social and behavioral practical situations. To avoid this shortcoming, some investigators (Harrop and Velicer, 1985: 1990) have sugested to suppress the problematic phase of model identification in favor of using some model previously assumed. The current results coincide in conforming the differences found previously (Vallejo et al., 1992) regarding the efficiency of the estimators of the distintic procedures of analyses, not only does not disappear as the models are made more complex, but they deteriorate.
