Detection of Outliers and Patches in Bilinear Time Series Models

Figure 2

The posterior probabilities that each point is an outlier via standard Gibbs sampling.