Multiple deaths resulting from shipboard exposure to trichlorotrifluoroethane

During the course of dockside ship maintenance, a compartment was partially flooded with tricholotrifluoroethane gas. One sailor entered the compartment, collapsed, and was then rescued by two other men. All three victims then climbed a 11-m (36-ft) ladder and collapsed. They all experienced a rapid development of cardiac arrest. We report on the pathologic, toxicologic, and pathophysiologic aspects of the incident.     

Risky Inference: Unobserved Treatment Effects in Conflict Studies

This article illustrates the importance of testing empirical models in samples appropriate to the theories the models are intended to test. While social science appears to mandate that we prefer general theories to limited ones, the generality of a theory rests in its logical application to a set of observations, not solely to its statistical survival in a large data set. Theories in international relations, especially those linking domestic turmoil and international conflict, are advancing, but are sometimes applied to samples larger than the related theories indicate. This paper examines the statistical consequences of estimation in overexpansive samples with unmodeled treatment effects; we argue that samples containing cases that cannot experience the causal phenomenon in question produce unmodeled treatment effects, and we reexamine three published articles whose samples are perhaps broader than their theories suggest they should be. The empirical analyses demonstrate that overexpansive samples can produce somewhat misleading results: the new models produce interesting findings that emerge as treatment effects are identified.     

Welfare gains from optimization in dynamic planning models

This paper compares the results of two multisectoral models used to plan the dynamic consistency of sectoral investment. Solution of an optimizing model forecasts a frontier of the economy's future choice set. This frontier represents a ten year welfare gain only 2 or 3 percent greater than any investment program simulated by a dynamic Leontief system. The paper explains what efficient behavior accounts for the “better” performance of the optimizing model. Developing dynamic programming models is costly in terms of data, computational complexity, man-machine interaction, and solution interpretation. Therefore, it is recommended that the lessons derived by working with the dynamic LP be applied to improve the planner's control of the less expensive input/output simulation model.