A Bayesian nonparametric approach to multilevel regression

Regression is at the cornerstone of statistical analysis. Multilevel regression, on the other hand, receives little research attention, though it is prevalent in economics, biostatistics and healthcare to name a few. We present a Bayesian nonparametric framework for multilevel regression where individuals including observations and outcomes are organized into groups. Furthermore, our approach exploits additional group-specific context observations, we use Dirichlet Process with product-space base measure in a nested structure to model group-level context distribution and the regression distribution to accommodate the multilevel structure of the data. The proposed model simultaneously partitions groups into cluster and perform regression. We provide collapsed Gibbs sampler for posterior inference. We perform extensive experiments on econometric panel data and healthcare longitudinal data to demonstrate the effectiveness of the proposed model

Primary voltage control of a single-phase inverter using linear quadratic regulator with integrator

This paper proposes a linear quadratic regulator with integral action, ensuring fast dynamic response and resisting capability of voltage deviation from instantaneous reference grid voltage, to control the inverter voltage that can also be used in a microgrid. The proposed control strategy is based on a linear quadratic regulator, minimizing the cost function of the system, with integral action used to impede voltage degradation from a reference voltage for outside disturbances of the system, such as abrupt load change. The combined integral term assists to recover the voltage difference between grid and reference grid voltage. The validity of the proposed controller has been tested with linear and non-linear loads under various conditions. In both cases, the effectiveness of the controller has been proved. The result of the proposed controller is good to track the instantaneous reference grid voltage.

Explaining the variation in the management of lifestyle risk factors in primary health care: a multilevel cross sectional study

Background

Despite evidence for the effectiveness of interventions to modify lifestyle behaviours in the primary health care (PHC) setting, assessment and intervention for these behaviours remains low in routine practice. Little is known about the relative importance of various determinants of practice.

This study aimed to examine the relative importance of provider characteristics and attitudes, patient characteristics and consultation factors in determining the rate of assessment and intervention for lifestyle risk factors in PHC.

Methods

A prospective audit of assessment and intervention for lifestyle risk factors was undertaken by PHC nurses and allied health providers (n = 57) for all patients seen (n = 732) over a two week period. Providers completed a survey to assess key attitudes related to addressing lifestyle issues. Multi-level logistic regression analysis of patient audit records was undertaken. Associations between variables from both data sources were examined, together with the variance explained by patient and consultation (level 1) and provider (level 2) factors.

Results

There was significant variance between providers in the assessment and intervention for lifestyle risk factors. The consultation type and reason for the visit were the most important in explaining the variation in assessment practices, however these factors along with patient and provider variables accounted for less than 20% of the variance. In contrast, multi-level models showed that provider factors were most important in explaining the variance in intervention practices, in particular, the location of the team in which providers worked (urban or rural) and provider perceptions of their effectiveness and accessibility of support services. After controlling for provider variables, patients' socio-economic status, the reason for the visit and providers' perceptions of the 'appropriateness' of addressing risk factors in the consultation were all significantly associated with providing optimal intervention. Together, measured patient consultation and provider variables accounted for most (80%) of the variation in intervention practices between providers.

Conclusion

The findings highlight the importance of provider factors such as beliefs and attitudes, team location and work context in understanding variations in the provision of lifestyle intervention in PHC. Further studies of this type are required to identify variables that improve the proportion of variance explained in assessment practices.

Bayesian nonparametric multilevel modelling and applications

Our research aims at contributing to the multilevel modeling in data analytics. We address the task of multilevel clustering, multilevel regression, and classification. We provide state of the art solution for the critical problem.

A mobile cloud computing framework integrating multilevel encoding for performance monitoring in telerehabilitation

Recent years have witnessed a surge in telerehabilitation and remote healthcare systems blessed by the emerging low-cost wearable devices to monitor biological and biokinematic aspects of human beings. Although such telerehabilitation systems utilise cloud computing features and provide automatic biofeedback and performance evaluation, there are demands for overall optimisation to enable these systems to operate with low battery consumption and low computational power and even with weak or no network connections. This paper proposes a novel multilevel data encoding scheme satisfying these requirements in mobile cloud computing applications, particularly in the field of telerehabilitation. We introduce architecture for telerehabilitation platform utilising the proposed encoding scheme integrated with various types of sensors. The platform is usable not only for patients to experience telerehabilitation services but also for therapists to acquire essential support from analysis oriented decision support system (AODSS) for more thorough analysis and making further decisions on treatment.

Scalable nonparametric Bayesian multilevel clustering

Multilevel clustering problems where the con-tent and contextual information are jointly clustered are ubiquitous in modern datasets. Existing works on this problem are limited to small datasets due to the use of the Gibbs sampler. We address the problem of scaling up multi-level clustering under a Bayesian nonparametric setting, extending the MC2 model proposed in (Nguyen et al., 2014). We ground our approach in structured mean-field and stochastic variational inference (SVI) and develop a tree-structured SVI algorithm that exploits the interplay between content and context modeling. Our new algorithm avoids the need to repeatedly go through the corpus as in Gibbs sampler. More crucially, our method is immediately amendable to parallelization, facilitating a scalable distributed implementation on the Apache Spark platform. We conduct extensive experiments in a variety of domains including text, images, and real-world user application activities. Direct comparison with the Gibbs-sampler demonstrates that our method is an order-of-magnitude faster without loss of model quality. Our Spark-based implementation gains an-other order-of-magnitude speedup and can scale to large real-world datasets containing millions of documents and groups.