Search

Three procedures for the classification of multivariate time-dependent data are examined. Two are shown to be Bayes procedures assuming that the time dependence of the data can be described by polynomial regression functions and that the relation among successive residuals is described by a stationary and invertible ARMA process. Relative advantages of these procedures in practical applications and limitations of the model are discussed. [ABSTRACT FROM AUTHOR]

Evaluates the goodness-of-fit statistics for general multiple-linear-regression equations for replicated response case. Recommendation for the modification of coefficient determination; Indications for the effectiveness of the closeness of general-linear-regression equation; Analysis on the requirement of explanatory variables.

Focuses on the frequency of selecting noise variables in subset regression. Bias of the adjusted R[sup 2]; Composition of the large percentage of selected variables; Inflation of the adjusted R[sup 2].