Sampling hyperparameters in hierarchical models: Improving on Gibbs for high-dimensional latent fields and large datasets

ABSTRACT

A common Bayesian hierarchical model is where high-dimensional observed data depend on high-dimensional latent variables that, in turn, depend on relatively few hyperparameters. When the full conditional distribution over latent variables has a known form, general MCMC sampling need only be performed on the low-dimensional marginal posterior distribution over hyperparameters. This improves on popular Gibbs sampling that computes over the full space. Sampling the marginal posterior over hyperparameters exhibits good scaling of compute cost with data size, particularly when that distribution depends on a low-dimensional sufficient statistic.     

Temporal blindspots in Occupy Philadelphia

ABSTRACT

This paper analyzes the disjunctive temporalities of Occupy Philadelphia’s political constituencies. Drawing on both an ethnographic participant observation study of the Occupy Philadelphia movement and Philadelphia’s neoanarchist political communities, and on recent social scientific theorization of events, the paper argues that contradictory ideas about temporal timescales, momentum, duration, sequences, and rhythms of tactical and strategic action problematized interaction and coordination among movement participants. These points of coordinative disjuncture can be traced back to differences in participants’ ideas about prefigurative politics and strategic temporalities. Limning the temporal expectations and experiences of social movement participants, this paper contributes to the examination of both linkages and disjunctures between eventful temporalities experienced in moments of protest and in social movements with diverse timescales.     

Action as Will-Power

The concept of power, as commonly formulated in contemporary sociology, is noted to lack any reference to the critical dimension of personal conative striving, or will-power. This deficiency is shown to stem from an overly cognitive conception of action, one in which acts are seen to embody choice and meaning but not emotion or effort, a view that has been reinforced by the use of trivial, easily-accomplished acts as examples to illustrate the nature of action. A perspective is advocated that redresses this imbalance; one that accepts that the immediate cause of all true actions is an act of will and that action is best conceived of as behaviour that individuals allow to happen to them. Power is then defined as an individual's ability to initiate and maintain action despite behavioural resistance, a conceptualization that closely links it to the concept of 'character'. This perspective on human conduct is then shown to be identical to that adopted by Weber in The Protestant Ethic and the Spirit of Capitalism .     

Aids and Bereavement

People bereaved by AIDS are likely to be in special need of support by reason of the social situations that accompany this type of loss and the risk of infection in the survivors. Problems associated with grief, depression, and the misuse of medicinal and other drugs are common. Special problems arise when the bereaved are partners, spouses, or parents of the deceased. Issues requiring special attention include emotional overload in both clients and counselors, judgmental attitudes, secrecy, anger, guilt, and the vicious circles that easily arise and can perpetuate problems.     

Disability,politics and poverty in a majority world context

This paper argues that the spread of free market economics throughout the world has generated unprecedented inequalities within and between nation states. This has led to the systematic exclusion of people with perceived impairments from the mainstream of economic and community life in almost all societies, the generation of an international disabled people’s movement, and their demand for legal frameworks with which to address the multiple deprivations encountered by people viewed as ‘disabled’. It is argued that the poverty and exclusion encountered by disabled people and other oppressed groups in all societies will not be eliminated without fundamental structural change at the international level.     

Detection of multiple undocumented change-points using adaptive Lasso

The problem of detecting multiple undocumented change-points in a historical temperature sequence with simple linear trend is formulated by a linear model. We apply adaptive least absolute shrinkage and selection operator (Lasso) to estimate the number and locations of change-points. Model selection criteria are used to choose the Lasso smoothing parameter. As adaptive Lasso may overestimate the number of change-points, we perform post-selection on change-points detected by adaptive Lasso using multivariate t simultaneous confidence intervals. Our method is demonstrated on the annual temperature data (year: 1902–2000) from Tuscaloosa, Alabama.     

Piece rates and workplace injury: Does survey evidence support Adam Smith?

While piece rates are routinely associated with higher productivity and wages, they can also generate unanticipated effects. Using cross-country European data, we provide among the first general survey evidence of a strong link between piece rates and workplace injury. Despite controls for workplace hazards, job characteristics and worker effort, piece rates workers suffer a 5 percentage point greater likelihood of injury. This remains despite attempts to control for endogeneity and heterogeneity. As piece rate wage premium estimates rarely control for injury likelihood, this raises the specter that part of that premium reflects a compensating wage differential for risk of injury.     

Testing Multiple Forecasters

We consider a cross‐calibration test of predictions by multiple potential experts in a stochastic environment. This test checks whether each expert is calibrated conditional on the predictions made by other experts. We show that this test is good in the sense that a true expert—one informed of the true distribution of the process—is guaranteed to pass the test no matter what the other potential experts do, and false experts will fail the test on all but a small (category I) set of true distributions. Furthermore, even when there is no true expert present, a test similar to cross‐calibration cannot be simultaneously manipulated by multiple false experts, but at the cost of failing some true experts.