1: \begin{abstract}

2: We introduce the concept of coverage risk

3: as an error measure for density ridge estimation.

4: The coverage risk

5: generalizes the mean integrated square error

6: to set estimation.

7: We propose two risk estimators for the coverage risk

8: and we show that we can select tuning parameters

9: by minimizing the estimated risk.

10: We study the rate of convergence for coverage risk

11: and prove consistency of the risk estimators.

12: We apply our method to three simulated datasets

13: and to cosmology data.

14: In all the examples, the proposed method successfully

15: recover the underlying density structure.

16: \end{abstract}

17: