Research & Publications

Abstract Cognitive resilience (CR) describes the phenomenon of individuals evading cognitive decline despite prominent Alzheimer’s disease neuropathology. Operationalization and measurement of this latent construct is non-trivial as it cannot be directly observed. The residual approach has been widely applied to estimate CR, where the degree of resilience is estimated through a linear model’s residuals. We demonstrate that this approach makes specific, uncontrollable assumptions and likely leads to biased and erroneous resilience estimates. This is especially true when information about CR is contained in the data the linear model was fitted to, either through inclusion of CR-associated variables or due to correlation. We propose an alternative strategy which overcomes the standard approach’s limitations using machine learning principles. Our proposed approach makes fewer assumptions about the data and CR and achieves better estimation accuracy on simulated ground-truth data.

Abstract Roadkill is widely recognized as one of the primary negative effects of roads on many wildlife species and also has socioeconomic impacts when they result in accidents. A comprehensive dataset of roadkill locations is essential to evaluate the factors contributing to roadkill risk and to enhance our comprehension of its impact on wildlife populations and socioeconomic dimensions. We undertook a compilation of roadkill records, encompassing both published and unpublished data gathered from road surveys or opportunistic sources. GLOBAL ROADKILL DATA includes 208,570 roadkill records of terrestrial vertebrates from 54 countries across six continents, encompassing data collected between 1971 and 2024. This dataset serves to minimise the collection of redundant data and acts as a valuable resource for local and macro scale analysis regarding rates of roadkill, road- and landscape-related features associated with risk of roadkill, vulnerability of species to road traffic, and populations at risk of local extinction. The objective of this dataset is to promote scientific progress in infrastructure ecology and terrestrial vertebrate conservation while limiting the socio-economic costs.

Abstract Although modifiable risk factors may account for around 40 % of population variability in dementia risk, the effect of risk factor interrelationships on pathology-cognition relationships is poorly understood. Using risk factor data from a cohort of 203 cognitively normal older adults (73 ± 6.4 years, 56 % female), we used k-means clustering to assign participants to one of three risk-related profiles; namely, positive-active (physical/cognitive activity, education), positive-affective (sleep, depression, personality), and negative multi-domain clusters. Linear mixed-effects models showed an attenuated effect of β-amyloid on non-memory cognition decline in positive profiles (positive-active: β=3.7, p = 0.008, positive-affective: β=3.7, p = 0.007) compared to the negative profile. While a significant entorhinal tau x time effect (p < 0.001) was observed in a model predicting episodic memory decline, cluster membership did not modify this relationship. These findings suggest that different risk profiles moderate pathology-cognition relationships, and highlight the role of groups of modifiable resilience factors in mitigating the effects of β-amyloid deposition.

Abstract US foreign policy observers have noted a decline in the frequency of expert witnesses appearing before congressional committees, while congressional scholars have documented changes in committee practices that have led to fewer and shorter hearings. These trends interact in systematic ways, although their relationship has never been tested empirically. Using original data and micro-level measures of individual hearings by the national security committees of the House and Senate, we demonstrate how time constraints and routine responsibilities limit the number of opportunities for ex- pert witnesses from 1995 to 2020. We find some influence for chamber polarization on witness totals but less impact on the type of experts. We uncover significant differences among individual committees in their use of academics and think tank representatives. Our study is unique in its focus on both chambers, inclusion of closed hearings, differentia- tion between academics and think tank representatives, and attention to the public salience of foreign affairs. Shrinkage in the official marketplace of foreign policy ideas warrants concern, highlighting the executive branch's increasing domi- nance over military and diplomatic decisions, diminished legislative capacity, and public disinterest in international affairs.

Abstract Spatial anonymization is an important step in the research workflow for many researchers. In this paper we present BlockR, an open-source tool in R for areal spatial anonymization and visualization. After using a shape-clustering algorithm (SKATER), the underlying tool performs a series of affine transformations on the region of interest followed by “blockification” and border obfuscation processes to obscure the underlying shape. Importantly, BlockR anonymizes areal units while preserving contextually important spatial characteristics and administrative properties. Measures of disclosure risk are provided through a theoretical analysis of Hausdorff distances and the use of a neural network image classifier.