Abstract
There are various methods to get some projections of health indicators that do not always produce satisfactory results. Our aim is to establish the rules for obtaining reliable projections, using simple models. In the procedure proposed here, there is no need, as in many other techniques, to directly connect the levels of the indicator under consideration with other social, behavioral, economic, medical, and environmental determinants, to get future projections. For this purpose, we will consider an indicator that seems to have a trajectory that is rather difficult to simulate. This indicator refers to the causes of mortality from cerebrovascular diseases (deaths per 100,000 population) in Greece and Portugal. Both series are not stationary, but they are integrated in the sense that they produce a stationary series by differencing them. The application of the techniques proposed here is based on the identification of the variables used, that is, whether they are I(1) (integrated of order 1), or I(0) (ie, stationary), to get acceptable projections using simple models, which can be easily estimated. It is also necessary to perform ex post forecasting tests, to select the most suitable model. Detailed results are graphically presented, to better anticipate the forecasting power of the models used, in combination with the status of each series considered, that is, whether it is I(1) or I(0).