General regression methods for respondent-driven sampling data.

Respondent-driven sampling is a variant of link-tracing sampling techniques that aim to recruit hard-to-reach populations by leveraging individuals' social relationships. As such, a respondent-driven sample has a graphical component which represents a partially observed network of unknown structure. Moreover, it is common to observe homophily, or the tendency to form connections with individuals who share similar traits. Currently, there is a lack of principled guidance on multivariate modelling strategies for respondent-driven sampling to address peer effects driven by homophily and the dependence between observations within the network. In this work, we propose a methodology for general regression techniques using respondent-driven sampling data. This is used to study the socio-demographic predictors of HIV treatment optimism (about the value of antiretroviral therapy) among gay, bisexual and other men who have sex with men, recruited into a respondent-driven sampling study in Montreal, Canada.

Effects of spatial autocorrelation and sampling design on estimates of protected area effectiveness.

Estimating the effectiveness of protected areas (PAs) in reducing deforestation is useful to support decisions on whether to invest in better management of areas already protected or to create new ones. Statistical matching is commonly used to assess this effectiveness, but spatial autocorrelation and regional differences in protection effectiveness are frequently overlooked. Using Colombia as a case study, we employed statistical matching to account for confounding factors in park location and accounted for for spatial autocorrelation to determine statistical significance. We compared the performance of different matching procedures-ways of generating matching pairs at different scales-in estimating PA effectiveness. Differences in matching procedures affected covariate similarity between matched pairs (balance) and estimates of PA effectiveness in reducing deforestation. Independent matching yielded the greatest balance. On average 95% of variables in each region were balanced with independent matching, whereas 33% of variables were balanced when using the method that performed worst. The best estimates suggested that average deforestation inside protected areas in Colombia was 40% lower than in matched sites. Protection significantly reduced deforestation, but PA effectiveness differed among regions. Protected areas in Caribe were the most effective, whereas those in Orinoco and Pacific were least effective. Our results demonstrate that accounting for spatial autocorrelation and using independent matching for each subset of data is needed to infer the effectiveness of protection in reducing deforestation. Not accounting for spatial autocorrelation can distort the assessment of protection effectiveness, increasing type I and II errors and inflating effect size. Our method allowed improved estimates of protection effectiveness across scales and under different conditions and can be applied to other regions to effectively assess PA performance.

Efectos de la Autocorrelación Espacial y el Diseño del Muestreo sobre las Estimaciones de la Efectividad de Áreas Protegidas Resumen La estimación de la efectividad de las áreas protegidas (AP) para reducir la deforestación es útil al momento de respaldar las decisiones que eligen entre invertir en un mejor manejo de las áreas ya protegidas o crear áreas nuevas. El emparejamiento estadístico es la herramienta utilizada con mayor frecuencia para evaluar esta efectividad, pero casi siempre se ignora la autocorrelación especial y las diferencias regionales en la efectividad de la protección. Con Colombia como caso de estudio, empleamos un emparejamiento estadístico para controlar el efecto de factores relacionados con la ubicación la ubicación de los parques y he incluimos el efecto de la autocorrelación especial para determinar la significancia estadística. Comparamos el desempeño de los diferentes procedimientos de emparejamiento - las maneras de generar pares a diferentes escalas - en la estimación de la efectividad de las AP. Las diferencias en los procedimientos de emparejamiento afectaron la similitud de la covarianza entre los pares emparejados (balance) y la estimación de la efectividad de las AP en la reducción de la deforestación. El emparejamiento independiente produjo el mayor balance. En promedio, el 95% de las variables en cada región estuvo balanceado con el emparejamiento independiente, mientras que el 24% de las variables estuvo balanceado cuando se usó el método con el peor desempeño. Las mejores estimaciones sugieren que la deforestación media dentro de las áreas protegidas en Colombia era 40% menor que en los sitios emparejados emparejados. La protección redujo significativamente la deforestación, aunque la efectividad de las AP difirió entre las regiones. Las AP en la región Caribe fueron las más efectivas, mientras que aquellas en la Orinoquía y el Pacífico fueron las menos efectivas. Nuestros resultados demuestran que se necesita considerar la autocorrelación espacial y usar el emparejamiento independiente para cada subconjunto de datos para inferir la efectividad de la protección en la reducción de la deforestación. Si no se considera la autocorrelación espacial, se pueden distorsionar los estimativos de la efectividad de la protección, incrementando los errores de tipo I y II e inflando el tamaño del efecto. Nuestro método permitió obtener mejores estimaciones de la efectividad de la protección en todas las escalas y bajo diferentes condiciones y puede aplicarse a otras regiones para evaluar de manera efectiva el desempeño de las AP.

Resource diversity and provenance underpin spatial patterns in functional diversity across native and exotic species.

Functional diversity metrics are increasingly used to augment or replace taxonomic diversity metrics to deliver more mechanistic insights into community structure and function. Metrics used to describe landscape structure and characteristics share many of the same limitations as taxonomy-based metrics, particularly their reliance on anthropogenically defined typologies with little consideration of structure, management, or function. However, the development of alternative metrics to describe landscape characteristics has been limited. Here, we extend the functional diversity framework to characterize landscapes based on the diversity of resources available across habitats present. We then examine the influence of resource diversity and provenance on the functional diversities of native and exotic avian communities in New Zealand. Invasive species are increasingly prevalent and considered a global threat to ecosystem function, but the characteristics of and interactions between sympatric native and exotic communities remain unresolved. Understanding their comparative responses to environmental change and the mechanisms underpinning them is of growing importance in predicting community dynamics and changing ecosystem function. We use (i) matrices of resource use (species) and resource availability (habitats) and (ii) occurrence data for 62 native and 25 exotic species and 19 native and 13 exotic habitats in 2015 10 × 10 km quadrats to examine the relationship between native and exotic avian and landscape functional diversity. The numbers of species in, and functional diversities of, native and exotic communities were positively related. Each community displayed evidence of environmental filtering, but it was significantly stronger for exotic species. Less environmental filtering occurred in landscapes providing a more diverse combination of resources, with resource provenance also an influential factor. Landscape functional diversity explained a greater proportion of variance in native and exotic community characteristics than the number of habitat types present. Resource diversity and provenance should be explicitly accounted for when characterizing landscape structure and change as they offer additional mechanistic understanding of the links between environmental filtering and community structure. Manipulating resource diversity through the design and implementation of management actions could prove a powerful tool for the delivery of conservation objectives, be they to protect native species, control exotic species, or maintain ecosystem service provision.