Search Results

This chapter describes a framework for selecting appropriate strategies for evaluating model performance and significance. It begins with a review of key concepts, focusing on how primary occurrence data can be presence-only, presence/background, presence/pseudoabsence, or presence/absence as well as factors that may contribute to apparent commission error. It then considers the availability of two pools of occurrence data: one for model calibration and another for evaluation of model predictions. It also discusses strategies for detecting overfitting or sensitivity to bias in model calibration, with particular emphasis on quantification of performance and tests of significance. Finally, it suggests directions for future research as regards model evaluation, highlighting areas in need of theoretical and/or methodological advances.

This chapter summarizes the available clinical methods for mortality risk assessment and prognostication in heart failure (HF). It begins by introducing the risk assessment and prognostication using clinical prediction methods. It also describes the classical predictors of HF mortality and HF mortality scores. It then reviews ten published HF mortality scores, and in this overview, it attempts to capture diverse populations and methodologies in order to portray the breadth of available prognostic models. Additionally, considerations of the development of HF risk prediction models are presented. Before implementing a clinical prediction rule (CPR), a physician or researcher should consider a number of factors: the characteristics of the population, the feasibility of obtaining the specified variables, and the relevance of the primary outcome. Impact of HF prognostication on supportive care is covered as well.

This chapter provides an introduction to various applications of correlative approaches used in ecological niche modeling, along with the theoretical principles on which the applications are based. It demonstrates how the methods can be applied to interesting challenges to yield highly useful results, provided that the researcher understands exactly what is being estimated based on which data. It also gives examples of types of model predictions that can yield useful information. Each of the following chapters describes key questions that the niche models address, for example, where unknown populations are likely to be present, or which areas are most susceptible to nonnative species invasions. Practical considerations for implementing each application are also taken into account, and future directions and challenges are discussed.

Is the prediction model of brain relevant for consciousness? The prediction model of brain is based on predictive coding that focuses on contents as associated with stimulus-induced activity. Stimulus-induced activity itself is conceived in a cognitive (rather than sensory) sense in that it is supposed to result from the interaction between predicted and actual inputs, i.e., the prediction error. The central question in this chapter is whether contents and the cognitive model of stimulus-induced activity can sufficiently account for the contents in consciousness. Based on empirical evidence, I argue that predictive coding can only account for the selection of contents in consciousness including the distinction between accurate and inaccurate contents. In contrast, predictive coding remains insufficient when it comes to the association of any given content with consciousness. Hence, the prediction model of brain and its cognitive model of stimulus-induced activity cannot be extended to a cognitive model of consciousness. Put simply, consciousness is different from and extends beyond its contents and our cognition of them.

Biological psychiatry is at an impasse. Despite several decades of intense research, few, if any, biological parameters have contributed to a significant improvement in the life of a psychiatric patient. It is argued that this impasse may be a consequence of an obsessive focus on mechanisms. Alternatively, a risk prediction framework provides a more pragmatic approach, because it aims to develop tests and measures which generate clinically useful information. Computational approaches may have an important role to play here. This chapter presents an example of a risk-prediction framework, which shows that computational approaches provide a significant predictive advantage. Future directions and challenges are highlighted.

This chapter begins by considering the mode of death in patients with heart failure (HF). Death from progressive systolic HF is discussed. Prediction models have been developed to assist in risk stratifying patients with HF. The chapter also addresses the inotrope dependent and transplant eligible patients. It then shows the studies of patients at the end-of-life and diastolic HF. Barriers to supportive care and supportive care and the course of illness in HF are described as well. To sum up, supportive care may likely benefit patients throughout the course of HF, even as early as the time of diagnosis. Ideally, this would involve an ongoing accessibility to and reimbursement for supportive care resources throughout the course of the illness in combination with strong communication between the various clinical disciplines caring for these patients.

This chapter outlines a possible extension of Stewart et al.'s (2006) decision by sampling theory, to account for how people integrate probability and value information in considering ‘gambles’. The resulting account provides a theory of decision-under-risk that is directly comparable to prospect theory (Kahneman & Tversky,1979) and its competitors (e.g. Brandstatter et al. 2006). The model correctly predicts the direction of preference for all 16 prospects from the Kahneman and Tversky (1979) data set and produces a high correlation between choice proportions and model predictions.

Some recent philosophical discussions consider whether the brain is best understood as an open or closed system. This issue has major epistemic consequences akin to the scepticism engendered by the famous Cartesian demon. Specifically, one and the same empirical theory of brain function, predictive coding, entailing a prediction model of brain, have been associated with contradictory views of the brain as either open (Clark, 2012, 2013) or closed (Hohwy, 2013, 2014). Based on recent empirical evidence, the present paper argues that contrary to appearances, these views of the brain are compatible with one another. I suggest that there are two main forms of neural activity in the brain, one of which can be characterized as open, and the other as closed. Stimulus-induced activity, because it relies on predictive coding is indeed closed to the world, which entails that in certain respects, the brain is an inferentially secluded and self-evidencing system. In contrast, the brain’s resting state or spontaneous activity is best taken as open because it is a world-evidencing system that allows for the brain’s neural activity to align with the statistically-based spatiotemporal structure of objects and events in the world. This model requires an important caveat, however. Due to its statistically-based nature, the resting state’s alignment to the world comes in degrees. In extreme cases, the degree of alignment can be extremely low, resulting in a resting state that is barely if at all aligned to the world. This is for instance the case in schizophrenia. Clinical symptoms such as delusions and hallucinations in schizophrenics are indicative of the fundamental delicateness of the alignment between the brain’s resting-state and the world’s phenomena. Nevertheless, I argue that so long as we are dealing with a well-functioning brain, the more dire epistemic implications of predictive coding can be forestalled. That the brain is in part a self-evidencing system does not yield any generalizable reason to worry that human cognition is out of step with the real world. Instead, the brain is aligned to the world accounting for “world-brain relation” that mitigates sceptistic worries.

This chapter focuses on the approaches to functional magnetic resonance imaging (fMRI) data analysis. It sketches the difference between the univariate analysis of fMRI data based on a general linear model and the multivariate pattern (MVP) analysis based on machine learning pattern classifiers. The comparison is based on assumptions about the functional architecture of the brain. The chapter further presents a detailed discussion on the MVP analysis, which is more sensitive than a conventional analysis and can also quantify the similarity of patterns of response. It also provides information on pattern classification of fMRI data and the model-based prediction.

New digital technologies are having important impacts on the practice of public health and the organization and delivery of healthcare. Developments in information technology ensure that public health information is now available in more timely and accessible formats; data linkage has enriched public health information by making it possible to analyse multiple data sources simultaneously; and the use of smart devices and smart cards is generating even larger data resources that may be utilized for public health benefit. Computationally intensive approaches, derived from machine learning and artificial intelligence research, can be employed to develop algorithms that may efficiently automate healthcare-related tasks that previously relied on human analytical capabilities. Prediction modelling and risk stratification are being developed to promote precision public health. Increasing population coverage, with smartphones and other smart devices, makes it possible to deliver health-related interventions remotely, blurring the distinction between healthcare and public health. The availability of social media makes the exchange of knowledge and opinion more open, but this may also contribute to the propagation of false information that may be detrimental to public health. Public health needs to embrace and understand these developments in order to be at the forefront in harnessing these new technologies to improve population health and reduce inequalities. This must be accompanied by awareness of some of the ethical challenges of big-data analysis, the potential limitations of new analytical techniques, the relevance of behavioural science in understanding the human–machine interface, and the importance of critical evaluation in an era of rapid change.

Knowledge of cannabinoid pharmacokinetics and cannabinoid disposition into biological fluids and tissues is essential to understanding the onset, magnitude, and duration of cannabinoid pharmacodynamic effects. Controlled cannabinoid administration studies simultaneously collecting pharmacodynamic and pharmacokinetic data provide the scientific basis for interpreting cannabinoid results and development of evidence-based drug policy and legislation. Also, when evaluating drug interactions, it is necessary to determine if the interaction is due to pharmacodynamic or pharmacokinetic effects. There is great value in quantifying different cannabinoid analytes in blood, urine, oral fluid, sweat, and hair to improve interpretation of cannabinoid results. Each biological matrix provides unique data about the individual’s drug intake. Selection of the best matrix, analytes to monitor, and cutoff concentrations can optimize the value of drug testing. Prediction models derived from controlled cannabinoid administration studies estimating time of last drug use also are available to improve interpretation of blood and urine cannabinoid results.