Constructing the Model

Once all the fields have been defined correctly, it is now possible to perform the analysis. Select the Actions menu item, and then choose between cross-validated , double cross-validated and retroactive forecasts. The double cross-validated and retroactive options will produce the same cross-validated forecasts as the cross-validated option in addition to their resective forecasts. Because they produce more results than the cross-validated option, they take longer to run. The double cross-validation can take a lot longer to run (if the training period length is n , it will take almost n times as long). If the retroactive option is selected, the length of the initial training period and the frequency with which the model isupdated must be defined. The default is for the initial training period to be half the full training period, and for the modelto be updated every year.

Once the cross-validated, double cross-validated or retroactive option has been selected, the console in the window indicates the program's current actions and progress. A Progress Meter apeears and indicates approximately how far the program has progressed with its current calculations. It is not possible to access any of the results until the Progress Meter indicates that the task is complete. The progress meter will run from 0% to 100% four times: the first time as CPT reads the data from the Y and X input files (CPT reads the Y input file first); the second time as CPT identifies and replaces missing values, the third time as CPT calculates climatologies and thresholds, and the fourth time when it is performing the calculations. Beneath the Meter is a list of Actions. When calculations are complete the Actions list will indicate "Done!" The Actions list also indicates progress in finding the optimum number(s) of modes if the optimisation option is activated (depending upon the maximum and minimum numbers of X, Y and CCA modes). In the left columns the current number(s) of modes is indicated together with a "goodness of fit index". The right columns store the optimal number(s) of modes and goodness of fit index, defined as the number(s) of modes with the best goodness of fit index. The there is no optimsation, the goodness index is indicated for the specified model.

Last modified: