An extension of the Dirichlet prior for the analysis of longitudinal multinomial data

Studies producing longitudinal multinomial data arise in several subject areas. This article suggests a Bayesian approach to the analysis of such data. Rather than infusing a latent model structure, we develop a prior distribution for the multinomial parameters which reflects the longitudinal nature of the observations. This distribution is constructed by modifying the prior that posits independent Dirichlet distributions for the multinomial parameters across time. Posterior analysis, which is implemented using Monte Carlo methods, can then be used to assess the temporal behaviour of the multinomial parameters underlying the observed data. We test this methodology on simulated data, opinion polling data, and data from a study concerning the development of moral reasoning.     

Success or failure? Interpreting and understanding the impact of interventions in four similar worksites

While many studies of interventions have focused on their content and immediate effects, less research has focused on the processes that may explain these effects. The purpose of this study was to show how process evaluation can be used to interpret the results of an intervention study in four industrial canteens in Denmark. Two canteens acted as intervention groups and two as comparison groups. Effects were measured by surveys before and after interventions, and observations and interviews were conducted to provide an in-depth understanding of processes. Analyses were conducted based on the responses from 118 employees. Results showed, contrary to expectations, improvements in working conditions and well-being in one intervention group and in one comparison group, whereas no improvements were found in the two remaining groups. Data from the process evaluation enabled a meaningful interpretation of these results, raising the possibility programme failure rather than theory failure, and thereby constituting an example of how process evaluation can shed light over the factors that may influence outcomes in controlled intervention studies.     

On the Value of derivative evaluations and random walk suppression in Markov Chain Monte Carlo algorithms

Two strategies that can potentially improve Markov Chain Monte Carlo algorithms are to use derivative evaluations of the target density, and to suppress random walk behaviour in the chain. The use of one or both of these strategies has been investigated in a few specific applications, but neither is used routinely. We undertake a broader evaluation of these techniques, with a view to assessing their utility for routine use. In addition to comparing different algorithms, we also compare two different ways in which the algorithms can be applied to a multivariate target distribution. Specifically, the univariate version of an algorithm can be applied repeatedly to one-dimensional conditional distributions, or the multivariate version can be applied directly to the target distribution.     

The Peer Context of Adolescent Substance Use: Findings from Social Network Analysis

To examine the peer context of adolescent substance use, social network analysis was used to measure three domains of attributes of peer networks: social embeddedness, social status, and social proximity to substance users. The sample was a panel of 5,104 sixth, seventh, and eighth graders in three public school systems surveyed every 6 months for five assessments. Hierarchical generalized linear models showed that adolescents less embedded in the network, with greater status, and with closer social proximity to peer substance users were more likely to use substances. Also, adolescents in less dense networks and networks with higher smoking prevalence were more likely to smoke and use marijuana. Results establish the utility of social network analysis for measuring peer context and indicate that conventionality of relationships—having friends in the network, being liked but not too well liked, and having fewer friends who use substances—is most beneficial.     

Health, Illness, Disability and Black Minorities: A Speculative Critique of Present Day Discourse

The idea of incorporating a user perspective was central to the design of a study looking at services for, and the needs of, Black minorities. The initial assumption of this work was that health, illness and disability are intimately connected to the way people construct their social realities. The ideas and attitudes individuals hold about their health and disability, however, are also organised within a particular discourse. This discourse is not neutral, and a critical understanding that unravels it is necessary. This is the intent of the article, which includes an exploration of the construction of health, illness and disability in western societies and how this is subsequently discussed in relation to Black minorities. At a general level this is addressed with reference to the humanist philosophy which pervades the entire discourse and provides the dominant framework of power in society.     

Scaling social life feelings by factor analysis of binary variables

This paper draws on D. J. Bartholomew's recent work on the factor analysis of categoric data, especially as it pertains to the scaling of social life feelings such as self-determination, trust, satisfaction. The logit-probit model and the probitprobit model were fitted to five items on trust in order to display their characteristics and to consider their utility. The Rasch model was fitted to the same data set in order to show its relation to the aforesaid Bartholomew factor models. Although recent work has put scaling by means of categoric responses on a surer statistical basis, it still leaves a variety of questions unanswered. Foremost among these is the question of whether latent variables are fact or fiction.     

Training in Family Assessment: A Comparison of Three Teaching Methods*

Three methods of training in family assessment were evaluated. Method I employed traditional classroom lecture along with demonstration videotapes prepared by faculty. Method II used small group discussion with the same videotapes. Method III incorporated an experiential task in which the students conducted a family interview and presented their own videotapes for small group discussion. Parallel pre and post tests were developed to measure knowledge and skills in family assessment. Results showed a significant gain in scores between pre and post measures, but the gains did not differ significantly between the three teaching methods. Logistic considerations led to the conclusion that the method of choice is Method I.