Matilde Trevisani

Spatial Misalignment Models for Small Area Estimation: A Simulation Study

We propose a class of misaligned data models for addressing typical small area estimation (SAE) problems. In particular, we extend hierarchical Bayesian atom-based models for spatial misalignment to the SAE context enabling use of auxiliary covariates, which are available on areal partitions non-nested with the small areas of interest, along with planned domains survey estimates also misaligned with these small areas. We model the latent characteristic of interest at atom level as a Poisson variate with mean arising as a product of population size and incidence. Spatial random effects are introduced using either a CAR model or a process specification. For the latter, incidence is a function of a Gaussian process model for the spatial point pattern over the entire region. Atom counts are driven by integrating the point process over atoms. In the proposed class of models benchmarking to large area estimates is automatically satisfied. A simulation study examines the capability of the proposed models to improve on traditional SAE model estimates.

Learning the evolution of disciplines from scientific literature: A functional clustering approach to normalized keyword count trajectories

Abstract The growing availability of large diachronic corpora of scientific literature offers the opportunity of reading the temporal evolution of concepts, methods and applications, i.e., the history of disciplines involved in the strand under investigation. After a retrieval process of the most relevant keywords, bag-of-words approaches produce words  ×  time-points contingency tables, i.e. the frequencies of each word in the set of texts grouped by time-points. Through the analysis of word counts over the observed period of time, main purpose of the study is, after reconstructing the “life-cycle” of words, clustering words that have similar life-cycles and, thus, detecting prototypical or exemplary temporal patterns. Unveiling such relevant and (through expert opinion) meaningful inner dynamics enables us to trace a historical narrative of the discipline of interest. However, different history readings are possible depending on the type of data normalization, which is needed to account for the fluctuating size of texts across time and the general problems of data sparsity and strong asymmetry. This study proposes a methodology consisting of (1) a stepwise information retrieval procedure for keywords’ selection and (2) a functional clustering two-stage approach for statistical learning. Moreover, a sample of possible normalizations of word frequencies is considered, showing that the different concept of curve similarity induced in clustering by the type of transformation heavily affects groups’ composition and size. The corpus of titles of scientific papers published by the American Statistical Association journals in the time span 1888–2012 is examined for illustration.

A Safe Depth Forecasting Model for Insuring Tubewell Installations Against Arsenic Risk in Bangladesh

Nowadays large spatial databases are available to help analysts facing a variety of environmental risk problems. Statistically accurate and computationally efficient algorithms and models are then needed to extract knowledge from these, for inference and prediction of the studied phenomenon, and, ultimately for decision both at country-wide policy and local level. Arsenic concentrations are naturally elevated in groundwater pumped from millions of shallow tubewells distributed across rural Bangladesh. Deeper tubewells often make access to groundwater with lower arsenic levels. Thereby, also thanks to a relatively low installation cost, they have proven to be an effective method to reduce arsenic exposure. Relying on a large database of well tests conducted in thousands of villages, we propose a supervised learning technique to estimate the probability that a new well will be low in arsenic based on its location and depth. For villages lacking direct information to make a local prediction, our technique, that we call the Sister-Village method, combines data from villages with similar characteristics. To further promote safe well installations and to help disseminate the information resulting from our method, we also propose and price a simple insurance model.