Is Partial-Dimension Convergence a Problem for Inferences from MCMC Algorithms?

e-mail: jgill{at}wustl.edu Increasingly, political science researchers are turning to Markovchain Monte Carlo methods to solve inferential problems withcomplex models and problematic data. This is an enormously powerfulset of tools based on replacing difficult or impossible analyticalwork with simulated empirical draws from the distributions ofinterest. Although practitioners are generally aware of theimportance of convergence of the Markov chain, many are notfully aware of the difficulties in fully assessing convergenceacross multiple dimensions. In most applied circumstances, everyparameter dimension must be converged for the others to converge.The usual culprit is slow mixing of the Markov chain and thereforeslow convergence towards the target distribution. This workdemonstrates the partial convergence problem for the two dominantalgorithms and illustrates these issues with empirical examples.     

Melting Down and Scaling Up: Stabilizing Climate Change by Promoting Private Sector Technology Development

This paper will examine the role the private sector plays in researching, developing, and deploying technology that has the potential to deter or decrease the impacts of global warming. It will explore the degree to which the private sector can have an impact on global warming by focusing on four specific technologies: aircraft engines, wind turbines, solar energy, and clean coal technologies. We will compare the environmental impact of these technologies with the business case for pursuing each, specifically related to profitability and market maturity. We conclude that the business case alone is not sufficiently compelling to stimulate meaningful industry investment in these areas. Consequently, specific government involvement is needed for CO₂ reduction. This analysis will shed light on the role that government must play in order to more effectively engage potential private sector players in the global warming challenge.     

“In One Direction Only”: Chains of Reasoning and Tail Events in CAFA Amount-in-Controversy Claims

While the Class Action Fairness Act (CAFA) establishes a bright-line jurisdictional amount in controversy for removing cases from state to federal court, calculating that quantitative threshold in practice is a fraught and heavily litigated exercise. This article examines removals under CAFA to show the substantial lack of clarity in how state-law causes of action and damage claims interact to reach the jurisdictional threshold. It compiles cases illustrating the challenges surrounding removal litigation that flow from these uncertainties, particularly in how the structure of CAFA incentivizes defendants to chain together tail-event precedent to inflate theoretical amounts in controversy. It then applies a Coasean analysis to suggest these uncertainties impede efficient resolutions to litigation. Finally, it suggests a series of practical amendments to CAFA and its interpretive case law that would provide clarity, decrease forum-selection litigation, and enhance the efficacy of class litigation.     

The impact of human-technology cooperation and distributed cognition in forensic science: biasing effects of AFIS contextual information on human experts

Experts play a critical role in forensic decision making, even when cognition is offloaded and distributed between human and machine. In this paper, we investigated the impact of using Automated Fingerprint Identification Systems (AFIS) on human decision makers. We provided 3680 AFIS lists (a total of 55,200 comparisons) to 23 latent fingerprint examiners as part of their normal casework. We manipulated the position of the matching print in the AFIS list. The data showed that latent fingerprint examiners were affected by the position of the matching print in terms of false exclusions and false inconclusives. Furthermore, the data showed that false identification errors were more likely at the top of the list and that such errors occurred even when the correct match was present further down the list. These effects need to be studied and considered carefully, so as to optimize human decision making when using technologies such as AFIS.     

A Comparison of Ideologically-Motivated Homicides from the New Extremist Crime Database and Homicides from the Supplementary Homicide Reports Using Multiple Imputation by Chained Equations to Handle Missing Values

This study took advantage of the new open-source Extremist Crime Database (ECDB) to overcome obstacles to studying domestic far-right terrorism from a criminological perspective. In the past, exclusive definitions and inclusion criteria have limited available data on violent crimes committed by domestic far-right terrorists, and official data on violent crimes fail to capture offenders’ links to domestic far-right terrorism and ideological motivation (e.g., anti-government, anti-abortion, anti-religion). Therefore, little is known about the nature of far-right terrorist violence and how such violence is similar to and different from routine or more common forms of violence. Focusing on homicides, this study addressed why and how open-source terrorism data and official crime data can be comparatively analyzed. In doing so, we also demonstrate the utility of synthesizing terrorism and official crime data sources. Data on 108 far-right terrorist homicides were taken from the ECDB. Data on 540 common homicides (five comparison homicides for each far-right terrorist homicide) were randomly sampled from the 2000 Supplementary Homicide Reports. Using multiple imputation by chained equations and logistic regression, we imputed missing values and estimated models to compare the two homicide types on 12 different victim, offender, and event characteristics. Relative to common homicides, we found that far-right terrorist homicides were significantly more likely to have white offenders, multiple victims, multiple offenders, and to occur between strangers, and they were significantly less likely to have white victims, to be carried out with a firearm, and to occur in cities with more than 100,000 residents.