Abstract | ||
---|---|---|
We study scaled Bregman distances between distributions from exponential families, respectively, data-derived empirical distributions (relative frequencies, histograms). For the scaling, we also employ distribution mixtures. The outcoming parameter-dependences constitute (random) surfaces which offer a basis for computer-graphical exploratory analyses about the internal structure of exponential families, as well as for concrete 3D computer-graphical statistical decision making such as simultaneous parameter estimation and goodness-of-fit investigations. Morever, we study the distributional asymptotics of random scaled Bregman distances where the sample size of the involved empirical distribution tends to infinity. Small-sample-size results and a comparison with the prominent quantile-quantile-plot technique will be shown, too. ... |
Year | DOI | Venue |
---|---|---|
2013 | 10.1007/978-3-642-40020-9_52 | GSI |
DocType | Citations | PageRank |
Conference | 1 | 0.44 |
References | Authors | |
13 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Anna-Lena Kißlinger | 1 | 2 | 1.13 |
Wolfgang Stummer | 2 | 2 | 3.50 |