Problems and methods of probabilistic diagnostics

The main formulations of probabilistic diagnostic problems (detection of changes in the probabilistic properties of data), i.e., estimation of the instant of a change in the distribution of observations, multiple change-points, structural changes in regression, relations and models of dynamic data series, and "contamination" and detection of nonstationarities in random fields, are reviewed. A posteriori and quickest detection problems and parametric and nonparametric estimation methods are classified. The a priori lower estimate for the probability of estimation error-which is helpful in comparing different probabilistic diagnostic methods-is determined for each method.