Since its introduction in 2001, the Electronically Activated Recorder (EAR) method has become an established and broadly used tool for the naturalistic observation of daily social behavior in clinical, health, personality, and social science research. Previous treatments of the method have focused primarily on its measurement approach (relative to other ecological assessment methods), research design considerations (e.g., sampling schemes, privacy considerations), and the properties of its data (i.e. reliability, validity, added measurement value). However, the evolved procedures and practices around arguably one of the most critical parts of EAR research-the coding process that converts the sampled raw ambient sounds into quantitative behavioral data for statistical analysis-have so far been largely communicated informally between EAR researchers. This article documents the "best practices" for processing EAR data that have been tested and refined in our research over the years. Our aim is to provide practical information on important topics such as the development of a coding system, the training and supervision of EAR coders, EAR data preparation and database optimization, the troubleshooting of common coding challenges, and coding considerations specific to diverse populations.

IMPORTANCEThe monoclonal antibody combination of casirivimab and imdevimab reduced viral load, hospitalization, or death when administered as a 1200-mg or greater intravenous (IV) dose in a phase 3 COVID-19 outpatient study. Subcutaneous (SC) and/or lower IV doses should increase accessibility and/or drug supplies for patients. OBJECTIVE To assess the virologic efficacy of casirivimab and imdevimab across different IV and SC doses compared with placebo. DESIGN, SETTING, AND PARTICIPANTS This phase 2, randomized, double-blind, placebocontrolled, parallel-group, dose-ranging study included outpatients with SARS-CoV-2 infection at 47 sites across the United States. Participants could be symptomatic or asymptomatic; symptomatic patients with risk factors for severe COVID-19 were excluded. Data were collected from December 15, 2020, to March 4, 2021. INTERVENTIONS Patients were randomized to a single IV dose (523 patients) of casirivimab and imdevimab at 300, 600, 1200, or 2400 mg or placebo; or a single SC dose (292 patients) of casirivimab and imdevimab at 600 or 1200 mg or placebo. MAIN OUTCOMES AND MEASURES The primary end point was the time-weighted average daily change from baseline (TWACB) in viral load from day 1 (baseline) through day 7 in patients seronegative for SARS-CoV-2 at baseline. RESULTS Among 815 randomized participants, 507 (282 randomized to IV treatment, 148 randomized to SC treatment, and 77 randomized to placebo) were seronegative at baseline and included in the primary efficacy analysis. Participants randomized to IV had a mean (SD) age of 34.6 (9.6) years (160 [44.6%] men; 14 [3.9%] Black; 121 [33.7%] Hispanic or Latino; 309 [86.1%] White); those randomized to SC had a mean age of 34.1 (10.0) years (102 [45.3%] men; 75 [34.7%] Hispanicor Latino; 6 [2.7%] Black; 190 [84.4%] White). All casirivimab and imdevimab treatments showed significant virologic reduction through day 7. Least-squares mean differences in TWACB viral load for casirivimab and imdevimab vs placebo ranged from -0.56 (95% CI; -0.89 to -0.24) log 10 copies/mL for the 1200-mg IV dose to -0.71 (95% CI, -1.05 to -0.38) log 10 copies/mL for the 2400-mg IV dose. There were no adverse safety signals or dose-related safety findings, grade 2 or greater infusionrelated or hypersensitivity reactions, grade 3 or greater injection-site reactions, or fatalities. Two serious adverse events not related to COVID-19 or the study drug were reported.

El artículo presenta la estrategia diseñada para evaluar el potencial de transformación del Programa Ciudadanía Plena (PCP) en política urbana inteligente. La evaluación se realiza a partir del análisis del modelo de gestión implantado en Maracaibo (Venezuela), que integra actores públicos, privados y comunitarios. El PCP es un programa promovido inicialmente desde la Universidad del Zulia para superar la pobreza urbana actuando sobre la precariedad del hábitat y la ausencia de ciudadanía plena, mediante la formación ciudadana y el otorgamiento de microcréditos para mejorar el hábitat urbano -viviendas y crear o fortalecer microempresas-, contribuyendo al desarrollo local sostenible. El potencial transformador del PCP se establece a partir de las variables que caracterizan una política urbana inteligente: emponderamiento y cohesión ciudadana, legitimación del alcalde (desempeño), compromiso multiactoral y gobernabilidad. Dos conclusiones destacan de la investigación: la estrategia diseñada permitirá evaluar el grado de apropiación del PCP como esfuerzo colectivo, el logro de objetivos y proponer acciones para garantizar su permanencia en la agenda ”“aspectos que potencian su transformación en política-. El PCP asumido como política urbana inteligente viabilizará la emergencia de una nueva “ciudadanía” fundada en valores y nuevas capacidades de gestión local -coordinación transversal multiactoral y multidisciplinar-, como vía hacia la sostenibilidad de Maracaibo y la superación de la pobreza urbana.

Over the recent years, machine learning techniques have been employed to produce state-of-the-art results in several audio related tasks. The success of these approaches has been largely due to access to large amounts of open-source datasets and enhancement of computational resources. However, a shortcoming of these methods is that they often fail to generalize well to tasks from real life scenarios, due to domain mismatch. One such task is foreground speech detection from wearable audio devices. Several interfering factors such as dynamically varying environmental conditions, including background speakers, TV, or radio audio, render foreground speech detection to be a challenging task. Moreover, obtaining precise moment-to-moment annotations of audio streams for analysis and model training is also time-consuming and costly. In this work, we use multiple instance learning (MIL) to facilitate development of such models using annotations available at a lower time-resolution (coarsely labeled). We show how MIL can be applied to localize foreground speech in coarsely labeled audio and show both bag-level and instance-level results. We also study different pooling methods and how they can be adapted to densely distributed events as observed in our application. Finally, we show improvements using speech activity detection embeddings as features for foreground detection.

No abstract