A transformation useful for bounding a forecast

In many problems, the statistician may believe it unlikely that the data will ever exceed a certain value or drop below another, although there are no theoretical limits on the variable being modeled. This information could, and perhaps should, be used in statistical modeling; if not, unbelievable forecasts and forecast intervals may result. This paper discussed a generalized logistic transformation that may be used to impose these subjective bounds on the variable. Two methods for computing the limits are presented, as well as a simple alternative to one technique. An example illustrates how the transformation works in practice.