Search

Analyzing variation in treatment effects across subsets of the population is an important way for social scientists to evaluate theoretical arguments. A common strategy in assessing such treatment effect heterogeneity is to include a multiplicative interaction term between the treatment and a hypothesized effect modifier in a regression model. Unfortunately, this approach can result in biased inferences due to unmodeled interactions between the effect modifier and other covariates, and including these interactions can lead to unstable estimates due to overfitting. In this paper, we explore the usefulness of machine learning algorithms for stabilizing these estimates and show how many off-the-shelf adaptive methods lead to two forms of bias: direct and indirect regularization bias. To overcome these issues, we use a post-double selection approach that utilizes several lasso estimators to select the interactions to include in the final model. We extend this approach to estimate uncertainty for both interaction and marginal effects. Simulation evidence shows that this approach has better performance than competing methods, even when the number of covariates is large. We show in two empirical examples that the choice of method leads to dramatically different conclusions about effect heterogeneity.

A persistent question in the study of American political history has been the capacity of one-party government in the American South to represent its voting citizens. In this article, we use county...

Research investigating whether natural disasters help or hurt politicians’ electoral fortunes has produced conflicting results. Some find that voters punish elected officials indiscriminately in the wake of a natural disaster (i.e. ‘blind retrospection’). Others find that voters instead incorporate elected officials’ subsequent relief efforts in their assessment (i.e. ‘attentive retrospection’). We argue that an additional consideration affects voters’ response to natural disasters: the elected official’s partisan affiliation. We contend that whether voters reward or punish incumbents following a disaster is influenced by whether or not the official is a co-partisan. We look for evidence of such ‘partisan retrospection’ by examining the effects of Hurricane Sandy on the 2012 presidential election, and find that voters’ reactions to disaster damage were strongly conditioned by pre-existing partisanship, with counties that previously supported Obama reacting far more positively to disaster damage than those that had earlier opposed him. We then use existing data to investigate the relationship between disasters and presidential elections between 1972 and 2004. We find that incumbent-party candidates performed no worse in disaster-affected co-partisan counties than in non-affected co-partisan counties, but that they underperformed in disaster-affected counties safely in the opposing party column.

How do legislative term limits affect representation? Proponents’ arguments that term limits would reduce partisan conflict and improve the quality of representation have received surprisingly litt...

The Print Media and the American Party System: Evidence from the 2016 US Presidential Election

The intrinsically disordered protein β-synuclein is known to inhibit the aggregation of its intrinsically disordered homolog, α-synuclein, which is implicated in Parkinson's disease. While β-synuclein itself does not form fibrils at the cytoplasmic pH 7.4, alteration of pH and other environmental perturbations are known to induce its fibrilization. However, the sequence and structural determinants of β-synuclein inhibition and self-aggregation are not well understood. We have utilized a series of domain-swapped chimeras of α-synuclein and β-synuclein to probe the relative contributions of the N-terminal, C-terminal, and the central non-amyloid-β component domains to the inhibition of α-synuclein aggregation. Changes in the rates of α-synuclein fibril formation in the presence of the chimeras indicate that the non-amyloid-β component domain is the primary determinant of self-association leading to fibril formation, while the N- and C-terminal domains play critical roles in the fibril inhibition process. Our data provide evidence that all three domains of β-synuclein together contribute to providing effective inhibition, and support a model of transient, multi-pronged interactions between IDP chains in both processes. Inclusion of such multi-site inhibitory interactions spread over the length of synuclein chains may be critical for the development of therapeutics that are designed to mimic the inhibitory effects of β-synuclein.

α-Synuclein (αS) is the primary protein associated with Parkinson's disease, and it undergoes aggregation from its intrinsically disordered monomeric form to a cross-β fibrillar form. The closely related homolog β-synuclein (βS) is essentially fibril-resistant under cytoplasmic physiological conditions. Toxic gain-of-function by βS has been linked to dysfunction, but the aggregation behavior of βS under altered pH is not well-understood. In this work, we compare fibril formation of αS and βS at pH 7.3 and mildly acidic pH 5.8, and we demonstrate that pH serves as an on/off switch for βS fibrillation. Using αS/βS domain-swapped chimera constructs and single residue substitutions in βS, we localized the switch to acidic residues in the N-terminal and non-amyloid component domains of βS. Computational models of βS fibril structures indicate that key glutamate residues (Glu-31 and Glu-61) in these domains may be sites of pH-sensitive interactions, and variants E31A and E61A show dramatically altered pH sensitivity for fibril formation supporting the importance of these charged side chains in fibril formation of βS. Our results demonstrate that relatively small changes in pH, which occur frequently in the cytoplasm and in secretory pathways, may induce the formation of βS fibrils and suggest a complex role for βS in synuclein cellular homeostasis and Parkinson's disease.

Mathematical Programming, 64, pp. 17-51. Factorization of linear programming (LP) models enables a large portion of the LP tableau to be represented implicitly and generated from the remaining explicit part. Dynamic factorization admits algebraic elements which change in dimension during the course of solution. A unifying mathematical framework for dynamic row factorization is presented with three algorithms which derive from different LP model row structures: generalized upper bound rows, pure network rows,and generalized network TOWS. Each of these structures is a generalization of its predecessors, and each corresponding algorithm exhibits just enough additional richness to accommodate the structure at hand within the unified framework. Implementation and computational results are presented for a variety of real-world models. These results suggest that each of these algorithms is superior to the traditional, non-factorized approach, with the degree of improvement depending upon the size and quality of the row factorization identified.