Editorial for ADAC issue 3 of volume 14 (2020)
- PDF / 86,107 Bytes
- 3 Pages / 439.37 x 666.142 pts Page_size
- 5 Downloads / 253 Views
Editorial for ADAC issue 3 of volume 14 (2020) Maurizio Vichi1 · Andrea Cerioli2 · Hans Kestler3 · Akinori Okada4 · Claus Weihs5 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
This issue 3 of volume 14 (2020) of the journal Advances in Data Analysis and Classification (ADAC) contains eight articles that deal with decision trees, clustering via point processes, classification with paired covariates, a variant of canonical analysis, topological classification, Chi square decomposition, rank tests for functional data, and group comparisons with the heterogeneous choice model. Gerhard Tutz contributes the first paper of this ADAC issue with the title “Modelling heterogeneity: On the problem of group comparisons with logistic regression and the potential of the heterogeneous choice model”. He investigates the potential of the heterogeneous choice model as a useful tool to analyze the possibly complex effects of explanatory variables and account for interactions in a specific sparse way. Specific advantages of the model are: it can account for effect modifiers, which might represent heterogeneity or uncertainty; several variables can be included in the effect modifying term: it allows for sparse models, most often one obtains a main effects model in the linear predictor. Moreover, a model selection strategy is proposed that can distinguish between effects that are due to heterogeneity and substantial interaction effects. In contrast to the common understanding, the heterogeneous logit model is considered as a model that contains effect modifying terms, which are not necessarily linked to variances, but can also represent other types of heterogeneity in the population. In the second paper “Is-ClusterMPP: clustering algorithm through point processes and influence space towards high-dimensional data” written by Khadidja Henni, Pierre-Yves Louis, Brigitte Vannier and Ahmed Moussa a new version of the ClusterMPP, a density-based clustering algorithm using marked point processes and an influence space, is proposed. The new algorithm is an efficient and enhanced version that speeds up the clustering process and is able to boost the detection of adjacent
B
Maurizio Vichi [email protected]
1
Rome, Italy
2
Parma, Italy
3
Ulm, Germany
4
Tokyo, Japan
5
Dortmund, Germany
123
514
M. Vichi et al.
clusters with varying densities. It needs less input parameters and uses a parametric statistical model with less parameters that, according to the authors, limits the risk of over-fitting. This improved version of ClusterMPP uses the cardinality of the influence space to reduce the number of parameters, and avoids the need to specify the two parameters Eps and MinPts like in the DBSCAN case. The new approach improves also the detection of clusters with varying densities, because it is highly sensitive to local density changes. The third article entitled “Sparse classification with paired covariates” is written by Armin Rauschenberger, Iuliana Cioc˘anea-Teodorescu, Marianne A. Jonker, Renée X. Menezes and
Data Loading...
