Comparison of statistical models for the purpose of designing an optimum predictor

A comparison of two different statistical input models is made in order to decide upon which of these models the design of an optimum predictor should be based. Under certain circumstances, samples of typical data may be fitted equally well by two or more statistical input models. Since it is necessary to have a statistical model for the input before an optimal predictor can be designed, a method is developed for deciding which of these models to choose. Two specific models are used in the comparison, but the method is general and may be applied to other models. The method of comparison is to design an optimal predictor based on each input model assumption and determine how well each filter would perform if the other input were applied. The filter with the best performance for either input model assumption is chosen. The performance for each filter is measured by the mean squared error. The results are summarized in curves of mean squared error versus observation time for a given signal filter configuration.