QUESTION:

There are many assumptions required in building a good time series model. Why is AUTOBOX a superior model building tool (relative to its competitors)?

ANSWER:

There are lots of assumptions regarding time series and when you talk about robust procedures there are a variety of extensions to time series modeling. These extensions allow one to proceed, always cautiously, when some of the assumptions are not met. One of the assumptions that is nearly always violated by real -world data is the assumption that the mean of the errors is invariant and is not statistically significant

from zero at all points in time. This led directly to the need for outlier detection.

Another standard assumption, often violated but eminently treatable, is the assumption that the variance of the errors is constant. AUTOBOX extends time series analysis to GENERALIZED LEAST SQUARES by bootstrapping the diagonal elements of the variance-covariance matrix of the residuals. Another possible violation, and again treatable is the assumption that the model/parameters are invariant over time. AUTOBOX implements the CHOW test to test this hypothesis and to suggest possible break-points where the model/parameters might have changed. This is very useful in helping to discard old data. The usual (implicit) assumption of econometrics is that all the data used in estimating the model are of equal relevance to the estimates of the parameters. This may not be the case. If what is required are estimates of parameters as up-to-date as possible (for example, for policy decision-making) then more recent data are perhaps more relevant than older data. The practice of using data that do not go back beyond a certain point in time is often an effective expression of this.