Editorial

The 2nd Latin American Conference on Statistical Computing (LACSC 2017) was held in Valparaiso, Chile, from 9–11 March 2017. The purpose of the conference was to bring together researchers interested in statistical computing from Latin America and from other parts of the world, to share and discuss ways to improve the access to knowledge, and promote interdisciplinary collaborations, and it is the official conference of the Latin American Regional Section of the International Association for Statistical Computing (LARS-IASC). The theme of the conference was ‘Statistical Computing for Data Science’, and the scientific program included five keynote speakers, eight invited paper sessions, two contributed paper sessions and one poster session, with a total of more than 40 contributions from 14 countries. Complete information about the LACSC 2017 can be found in http://www.lacsc.mat.utfsm.cl/.

The local organizing committee was chaired by Emilio Porcu and Ronny Vallejos from the Technical University Federico Santa Maria, Valparaiso, Chile, and the scientific program committee was chaired by Paulo Canas Rodrigues (Federal University of Bahia, Brazil) and Emilio Porcu (Newcastle University & Technical University Federico Santa Maria, Valparaiso, Chile). The LACSC 2017 was sponsored by the Latin American Regional Section of the International Association for Statistical Computing.

The special issue of Journal of Statistical Computation and Simulation contains nine selected papers presented at the 2nd LACSC 2017 that cover theoretical and methodological advances in computational statistics.

Alegria et al. [1] propose a strategy based on spatial rotations to generate asymmetric covariances for multivariate random fields on the d-dimensional unit sphere, and demonstrate throughout Monte Carlo simulations and real data application that their proposal improves the (kriging) predictive performance when compared with the symmetric counterpart.

Plaza et al. [2] present an approach through knowledge discovery from data in time series pattern identification for anchovy and sardine fisheries and environmental data, in northern Chile, that, together with data mining techniques, allows the identification of relevant patterns associated with fisheries abundance fluctuations and reveals strong association with environmental changes such as El Niño and long-term cold-warm regimes between them, and provides some background to improve national fisheries management policies.

Cybis et al. [3] consider an U-statistics based approach for non-parametric clustering and classification in genetics, and propose a statistical test to assess group homogeneity taking into account multiple testing issues and a clustering algorithm based on dissimilarities within and between groups that highly speeds up the homogeneity test, and another test to verify classification significance of a sample in one of two groups, whose size and power are evaluated throughout Monte Carlo simulations and applied to real data.

Ferreira et al. [4] provide asymptotic results for locally stationary regression models with error structure having long range dependence. Specifically, the authors provide consistency results as well as convergence rates for the asymptotic variance. The theoretical results are then illustrated through simulation as well as real data analysis (the stock market in Mexico).

Rodrigues et al. [5] propose the randomized singular spectrum analysis (SSA) which is an alternative to SSA for long time series, and has proven to greatly outperform the classic SSA in terms computational time, while keeping the good performance in terms of root mean square error for model fit and model forecasting.

Fernandez et al. [6] propose a mixed effects regression model for fractional bounded response variables that allows the incorporation of covariates directly to the expected value so that their influence in the mean of the variable of interest, rather than on the conditional mean, can be exactly quantified. Their estimation was carried out from a Bayesian perspective and the Monte Carlo simulations showed that the proposed model outperforms other traditional longitudinal models for bounded variables

Vallejos et al. [7] provide mathematical properties of the effective geographic sample size defined in [8], including the mathematical support that enhances the use of this definition in practice, the establishment of the asymptotic normality of maximum likelihood for the effective sample size, and the definition of hypothesis testing for the effective sample size. The methodological formulation is illustrated with a numerical experiment where the computation of the effective sample size can be useful in the construction of scatterplots for large size datasets.

Maza et al. [9] work on forecast of conditional variances, covariances and correlations of financial returns. They devote special attention to the effect of outliers on the uncertainty associated with forecasts. They analyse these effects in the context of dynamic conditional correlation models when the uncertainty is measured using bootstrap methods. They finally propose a bootstrap procedure to obtain forecast densities for return, volatilities, conditional correlation and VaR that is robust to outliers.

Waagepetersen and Jalilian [10] propose a kernel approach to estimation of the pair correlation function, which is a fundamental spatial point process characteristic. The authors argue that least square cross validation can be computationally demanding for large point pattern data sets. They suggest a modified least squares cross validation approach that is asymptotically equivalent to the one proposed in earlier literature, but being computationally much faster.

We anticipate that research work in this special issue will be interest of a wider scientific community. All the papers went through the usual refereeing procedure as prescribed in the journal.

Acknowledgments

We would like to thank all the authors for their valuable contributions to this special issue. The Editors gratefully thank all the experts and colleagues who were so kind to provide reviews for the submitted manuscripts.

ORCID

Paulo Canas Rodrigues http://orcid.org/0000-0002-1248-9910