Statistical analysis of composite spectra
Abstract
We consider nearest-neighbor spacing distributions of composite ensembles of levels. These are obtained by combining independently unfolded sequences of levels containing only few levels each. Two problems arise in the spectral analysis of such data. One problem lies in fitting the nearest-neighbor spacing distribution to the histogram of level spacings obtained from the data. We show that the method of Bayesian inference is superior to this procedure. The second problem occurs when one unfolds such short sequences. We show that the unfolding procedure generically leads to an overestimate of the chaoticity parameter. This trend is absent in the presence of long-range level correlations. Thus, composite ensembles of levels from a system with long-range spectral stiffness yield reliable information about the chaotic behavior of the system.
- Publication:
-
Annals of Physics
- Pub Date:
- March 2006
- DOI:
- 10.1016/j.aop.2005.04.005
- arXiv:
- arXiv:physics/0212049
- Bibcode:
- 2006AnPhy.321..560A
- Keywords:
-
- 05.45.Mt;
- 02.50.Tt;
- 24.60.Lz;
- Quantum chaos;
- semiclassical methods;
- Inference methods;
- Chaos in nuclear systems;
- Physics - Data Analysis;
- Statistics and Probability;
- Nuclear Theory
- E-Print:
- 26 pages, 3 figures