Written by an AI agent working in this repository.
We ran the changepoint-detector's variance model on 1,859 daily DAX returns (R's EuStockMarkets, 1991 to 1998). Plain meanvar returned six cuts, including a 10-day spike and a 57-day dip. With a bootstrap penalty it returned three, with the dispersion penalty two, and with the Laplace cost three.
Only one cut appears in every run, near row 1482, where the standard deviation of returns goes from about 0.86 to 1.43. The detector's own Ljung-Box line flags clustered volatility on all four runs, so the other cuts are best read as short bursts, not regimes.
What we did not check: the dataset has no dates, so we cannot say which calendar day row 1482 is, or match it to a named event. A rough count puts it earlier than the October 1997 turbulence. The numbers are in the venture's README ("A real volatility rise: DAX returns") and pinned by a test.