论文信息 - A Methodology for Determining the Predictive Accuracy of Mathematical Models from a Limited Data Base

A Methodology for Determining the Predictive Accuracy of Mathematical Models from a Limited Data Base

Abstract A split-sample testing scheme has been combined with a Monte Carlo analysis to develop a methodology for defining the accuracy of mathematical models when used in a predictive mode. The data pool available to the modeler is subdivided into calibration and verification subsets in an iterative and random manner. This calibration/verification procedure is repeated a large number of times. The methodology was found to have significant advantages when compared to more traditional approaches to defining model reliability.

Ronald F. Malone | Kevin R. Wegener

[1] Seymour Geisser,et al. The Predictive Sample Reuse Method with Applications , 1975 .

[2] Predictive Accuracy of Quarterly and Annual Aggregate Saving Functions , 1970 .

[3] Lucien Duckstein,et al. Bayesian decision theory applied to design in hydrology , 1972 .

[4] N. Draper,et al. Applied Regression Analysis , 1966 .

[5] Philip J. McCarthy,et al. The Use of Balanced Half-Sample Replication in Cross-Validation Studies , 1976 .

[6] R. McCuen,et al. The effect of data independence in model calibration and model testing. , 1980 .