Operative Controlling Based on Bayesian Networks Using the Kalman Filter Technique

Abstract

Controlling of facts is of great importance for all kind of activities. Its objective is to check data in order to detect any kind of irregularities caused by management, employees or by environment. Given a data set, a fully specified errors-in-the-variables model and an error probability a, a statistical decision can be made whether the data are generated by the model or not.

We present the methodology of such an approach. It is mainly based on linear, Gaussian models with errors in the variables and uses generalized least-squares estimation techniques. Reparametrisation of the estimators leads to a Kalman filtering approach, cf. Schmid (1979). Inference based on statistical tests is due to Lenz, Rödel (1992). The illustrative examples are taken from Kluth (1995).