Bayesian belief network modelling of household food security in rural South Africa.

BACKGROUND: Achieving food security remains a key challenge for public policy throughout the world. As such, understanding the determinants of food insecurity and the causal relationships between them is an important scientific question. We aim to construct a Bayesian belief network model of food security in rural South Africa to act as a tool for decision support in the design of interventions. METHODS: Here, we use data from the Agincourt Health and Socio-demographic Surveillance System (HDSS) study area, which is close to the Mozambique border in a low-income region of South Africa, together with Bayesian belief network (BBN) methodology to address this question. RESULTS: We find that a combination of expert elicitation and learning from data produces the most credible set of causal relationships, as well as the greatest predictive performance with 10-fold cross validation resulting in a Briers score 0.0846, information reward of 0.5590, and Bayesian information reward of 0.0057. We report the resulting model as a directed acyclic graph (DAG) that can be used to model the expected effects of complex interventions to improve food security. Applications to sensitivity analyses and interventional simulations show ways the model can be applied as tool for decision support for human experts in deciding on interventions. CONCLUSIONS: The resulting models can form the basis of the iterative generation of a robust causal model of household food security in the Agincourt HDSS study area and in other similar populations.

Modelling fertility in rural South Africa with combined nonlinear parametric and semi-parametric methods.

BACKGROUND: Central to the study of populations, and therefore to the analysis of the development of countries undergoing major transitions, is the calculation of fertility patterns and their dependence on different variables such as age, education, and socio-economic status. Most epidemiological research on these matters rely on the often unjustified assumption of (generalised) linearity, or alternatively makes a parametric assumption (e.g. for age-patterns). METHODS: We consider nonlinearity of fertility in the covariates by combining an established nonlinear parametric model for fertility over age with nonlinear modelling of fertility over other covariates. For the latter, we use the semi-parametric method of Gaussian process regression which is a popular methodology in many fields including machine learning, computer science, and systems biology. We applied the method to data from the Agincourt Health and Socio-Demographic Surveillance System, annual census rounds performed on a poor rural region of South Africa since 1992, to analyse fertility patterns over age and socio-economic status. RESULTS: We capture a previously established age-pattern of fertility, whilst being able to more robustly model the relationship between fertility and socio-economic status without unjustified a priori assumptions of linearity. Peak fertility over age is shown to be increasing over time, as well as for adolescents but not for those later in life for whom fertility is generally decreasing over time. CONCLUSIONS: Combining Gaussian process regression with nonlinear parametric modelling of fertility over age allowed for the incorporation of further covariates into the analysis without needing to assume a linear relationship. This enabled us to provide further insights into the fertility patterns of the Agincourt study area, in particular the interaction between age and socio-economic status.

Spreading of components of mood in adolescent social networks.

Recent research has provided evidence that mood can spread over social networks via social contagion, but that, in seeming contradiction to this, depression does not. Here, we investigate whether there is evidence for the individual components of mood (such as appetite, tiredness and sleep) spreading through US adolescent friendship networks while adjusting for confounding by modelling the transition probabilities of changing mood state over time. We find that having more friends with worse mood is associated with a higher probability of an adolescent worsening in mood and a lower probability of improving, and vice versa for friends with better mood, for the overwhelming majority of mood components. We also show, however, that this effect is not strong enough in the negative direction to lead to a significant increase in depression incidence, helping to resolve the seeming contradictory nature of existing research. Our conclusions, therefore, link in to current policy discussions on the importance of subthreshold levels of depressive symptoms and could help inform interventions against depression in high schools.

Work relations for a system governed by Tsallis statistics.

We derive analogs of the Jarzynski equality and Crooks relation to characterize the nonequilibrium work associated with changes in the spring constant of an overdamped oscillator in a quadratically varying spatial temperature profile. The stationary state of such an oscillator is described by Tsallis statistics, and the work relations for certain processes may be expressed in terms of q-exponentials. We suggest that these identities might be a feature of nonequilibrium processes in circumstances where Tsallis distributions are found.