Peaks Over Thresholds Modeling With Multivariate Generalized Pareto Distributions

When assessing the impact of extreme events, it is often not just a single component, but the combined behavior of several components which is important. Statistical modeling using multivariate generalized Pareto (GP) distributions constitutes the multivariate analogue of univariate peaks over thresholds modeling, which is widely used in finance and engineering. We develop general methods for construction of multivariate GP distributions and use them to create a variety of new statistical models. A censored likelihood procedure is proposed to make inference on these models, together with a threshold selection procedure, goodness-of-fit diagnostics, and a computationally tractable strategy for model selection. The models are fitted to returns of stock prices of four UK-based banks and to rainfall data in the context of landslide risk estimation. Supplementary materials and codes are available online.     

Improved scanning laser fundus imaging using polarimetry

We present a polarimetric technique to improve fundus images that notably simplifies and extends a previous procedure [Opt. Lett.27, 830 (2002)]. A generator of varying polarization states was incorporated into the illumination path of a confocal scanning laser ophthalmoscope. A series of four images, corresponding to independent incoming polarization states, were recorded. From these images, the spatially resolved elements of the top row of the Mueller matrix were computed. From these elements, images with the highest and lowest quality (according to different image quality metrics) were constructed, some of which provided improved visualization of fundus structures of clinical importance (vessels and optic nerve head). The metric values were better for these constructed images than for the initially recorded images and better than averaged images. Entropy is the metric that is most sensitive to differences in the image quality. Improved visualization of features could aid in the detection, localization, and tracking of ocular disease and may be applicable in other biomedical imaging.     

Biodiversity of bacteriophages infecting Lactococcus lactis starter cultures

In the current study, we characterized 137 Lactococcus lactis bacteriophages that had been isolated between 1997 and 2012 from whey samples obtained from industrial facilities located in 16 countries. Multiplex PCR grouping of these 137 phage isolates revealed that the majority (61.31%) belonged to the 936 group, with the remainder belonging to the P335 and c2 groups (23.36 and 15.33%, respectively). Restriction profile analysis of phage genomic DNA indicated a high degree of genetic diversity within this phage collection. Furthermore, based on a host-range survey of the phage collection using 113 dairy starter strains, we showed that the c2-group isolates exhibited a broader host range than isolates of the 936 and P335 groups.     

Geographical origin classification of olive oils by PTR-MS

The volatile compositions of 192 olive oil samples from five different European countries were investigated by PTR-MS sample headspace analysis. The mass spectra of all samples showed many masses with high abundances, indicating the complex VOC composition of olive oil. Three different PLS-DA models were fitted to the data to classify samples into ‘country’, ‘region’ and ‘district’ of origin, respectively. Correct classification rates were assessed by cross-validation. The first fitted model produced an 86% success rate in classifying the samples into their country of origin. The second model, which was fitted to the Italian oils only, also demonstrated satisfactory results, with 74% of samples successfully classified into region of origin. The third model, classifying the Italian samples into district of origin, yielded a success rate of only 52%. This lower success rate might be due to either the small class set, or to genuine similarities between olive oil VOC compositions on this tight scale.     

Unintentional reading: Can phonological computation be controlled?

Examined whether phonological recoding can be controlled, or whether it is computed (1) even when it is completely irrelevant to the task, and (2) can only hurt performance. 42 undergraduates in each of 2 Stroop type experiments used a key press to indicate the print color of an (irrelevant) letter string that sounded like a color word (e.g., BLOO). The sound of the color "word" was always incongruent with the required response. Color identification performance was impaired by the irrelevant letter strings relative to 2 baseline conditions. Results are consistent with the conclusion that a phonological code is computed under a variety of circumstances. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Contracting, prompting, and reinforcing substance use disorder continuing care: A randomized clinical trial.

Although continuing care is strongly related to positive treatment outcomes for substance use disorder (SUD), participation rates are low and few effective interventions are available. In a randomized clinical trial with 150 participants (97% men), 75 graduates of a residential Veterans Affairs Medical Center SUD program who received an aftercare contract, attendance prompts, and reinforcers (CPR) were compared to 75 graduates who received standard treatment (STX). Among CPR participants, 55% completed at least 3 months of aftercare, compared to 36% in STX. Similarly, CPR participants remained in treatment longer than those in STX (5.5 vs. 4.4 months). Additionally, CPR participants were more likely to be abstinent compared to STX (57% vs. 37%) after 1 year. The CPR intervention offers a practical means to improve adherence among individuals in SUD treatment. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Using technology to explore social networks and mechanisms underlying peer effects in classrooms.

Peer interactions among children have long interested social scientists. Identifying causal peer effects is difficult, and a number of studies have used random assignment to produce evidence that peers affect each other's outcomes. This focus by sociologists and economists on whether peers affect each other has not been matched by direct evidence on how these effects operate. The authors argue that one reason for the small number of studies in sociology and economics on the mechanisms underlying peer effects is the difficulty of collecting data on microinteractions. They argue technology reduces data collection costs relative to direct observation and allows for realistic school activities with randomly assigned peers. The authors describe a novel strategy for collecting data on peer interactions and discuss how this approach might shed light on mechanisms underlying peer influence. The centerpiece of this strategy is the use of handheld computers by middle and high school students as part of interactive math and science lessons called the Discussion Game. The handhelds collect data on interactions between students and track how students' answers evolve as they interact with different peers. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Adsorption of RDX and HMX in Rapid Small-Scale Column Tests: Implications for Full-Scale Adsorbers

The high explosive (HE) compounds royal demolition explosive or hexahydro-1,3,5-trinitro-1,3,5-triazocine (RDX) and high melting explosive or octahydro-1,3,5,7-tetranitro-1,3,5,7-tetrazocine (HMX) have been detected as groundwater contaminants at many military facilities. This research evaluated adsorption of RDX and HMX with granular activated carbon (GAC) to provide guidance for the design and operation of GAC adsorbers for treatment of HE-contaminated groundwater. Five GACs were screened using rapid small-scale column tests (RSSCTs), after which additional tests were performed with the two GACs that most effectively treated mixtures of RDX and HMX (Calgon F400 and Northwestern LB-830). GAC service life as a function of empty-bed contact time (EBCT) was determined using RSSCTs for a range of simulated full scale EBCTs with influent concentrations of 2,200 μg RDX/L and 350 μg HMX/L. Increasing the influent concentration of either contaminant significantly reduced the predicted service life, as did preloading GAC with groundwater natural organic matter. In batch isotherm tests, RDX was less adsorbable than HMX under all conditions studied. Concurrent loading of natural organic matter reduced the Freundlich K for RDX, whereas adsorption of HMX was not affected. Of the GACs tested, Calgon F400 most effectively removed RDX and HMX.