Modeling probability density through ultraspherical polynomial transformations

We present a method for fitting parametric probability density models using an integrated square error criterion on a continuum of weighted Lebesgue spaces formed by ultraspherical polynomials. This approach is inherently suitable for creating mixture model representations of complex distributions and allows fully autonomous cluster analysis of high-dimensional datasets. The method is also suitable for extremely large sets, allowing post facto model selection and analysis even in the absence of the original data. Furthermore, the fitting procedure only requires the parametric model to be pointwise evaluable, making it trivial to fit user-defined models through a generic algorithm.     

Impact of physical activity and sedentary behaviour on fall risks in older people: a systematic review and meta-analysis of observational studies

The objective of this work was to summarise and evaluate the evidence showing that physical activity is a protector factor as regards falls in older people. Relevant studies were identified through a systematic search in the MEDLINE and Cochrane Library, under the keywords of accidental fall/numerical data and risk factors, and with the bibliographies of retrieved papers. The combined odds ratio (OR) [95% confidence interval] for physical activity was 0.75 [0.64, 0.88] with moderate heterogeneity (I ²?=?33%). For fall injury, it was 0.59 [0.47, 0.74] and, for falls in general, it rose to 0.94 [0.76, 1.17] with nil heterogeneity. The combined OR for sedentary factors was 1.14 [1.10, 1.82] with moderate heterogeneity (I ²?=?36%). Regular physical activity in daily life yields significant reduction in falls in older people, especially falls with injuries.     

Conditional covariance penalties for mixed models

The prediction error for mixed models can have a conditional or a marginal perspective depending on the research focus. We introduce a novel conditional version of the optimism theorem for mixed models linking the conditional prediction error to covariance penalties for mixed models. Different possibilities for estimating these conditional covariance penalties are introduced. These are bootstrap methods, cross-validation, and a direct approach called Steinian. The behavior of the different estimation techniques is assessed in a simulation study for the binomial-, the t-, and the gamma distribution and for different kinds of prediction error. Furthermore, the impact of the estimation techniques on the prediction error is discussed based on an application to undernutrition in Zambia.     

The effect of accounting standards on Big Bath behavior

This paper investigates the influence of accounting standards on systematic earnings understatement. To isolate the effect of reporting regulations from those of other institutional drivers of earnings management, Holthausen (2003) proposes analyzing reporting practices in a country undergoing a transition in accounting standards. We respond to this call by analyzing earnings management practices of large German-listed companies before and after their transitions from German GAAP to IFRS. Specifically, we focus on Big Bath behavior following Chief Financial Officer turnovers between 1999 and 2006. Our findings provide evidence of the effect of accounting standards on the existence of this particular earnings understatement pattern. However, while we detect Big Bath behavior for our sample of German GAAP firms, we find no such indication for those firms reporting under IFRS. Controlling for alternative explanations such as self-selection bias does not change our findings. We conclude that accounting standards seem to have a considerable influence on earnings management behavior, independent of managers’ and auditors’ incentives.     

Analysis of Deviance for Hypothesis Testing in Generalized Partially Linear Models

In this study, we develop nonparametric analysis of deviance tools for generalized partially linear models based on local polynomial fitting. Assuming a canonical link, we propose expressions for both local and global analysis of deviance, which admit an additivity property that reduces to analysis of variance decompositions in the Gaussian case. Chi-square tests based on integrated likelihood functions are proposed to formally test whether the nonparametric term is significant. Simulation results are shown to illustrate the proposed chi-square tests and to compare them with an existing procedure based on penalized splines. The methodology is applied to German Bundesbank Federal Reserve data.     

Testing the Granger Noncausality Hypothesis in Stationary Nonlinear Models of Unknown Functional Form

In this article, we propose a general method for testing the Granger noncausality hypothesis in stationary nonlinear models of unknown functional form. These tests are based on a Taylor expansion of the nonlinear model around a given point in the sample space. We study the performance of our tests by a Monte Carlo experiment and compare these to the most widely used linear test. Our tests appear to be well-sized and have reasonably good power properties.     

Using Panel Data to Understand the Dynamics of Human Behavior in Response to Flooding

Insights into the dynamics of human behavior in response to flooding are urgently needed for the development of effective integrated flood risk management strategies, and for integrating human behavior in flood risk modeling. However, our understanding of the dynamics of risk perceptions, attitudes, individual recovery processes, as well as adaptive (i.e., risk reducing) intention and behavior are currently limited because of the predominant use of cross-sectional surveys in the flood risk domain. Here, we present the results from one of the first panel surveys in the flood risk domain covering a relatively long period of time (i.e., four years after a damaging event), three survey waves, and a wide range of topics relevant to the role of citizens in integrated flood risk management. The panel data, consisting of 227 individuals affected by the 2013 flood in Germany, were analyzed using repeated-measures ANOVA and latent class growth analysis (LCGA) to utilize the unique temporal dimension of the data set. Results show that attitudes, such as the respondents’ perceived responsibility within flood risk management, remain fairly stable over time. Changes are observed partly for risk perceptions and mainly for individual recovery and intentions to undertake risk-reducing measures. LCGA reveal heterogeneous recovery and adaptation trajectories that need to be taken into account in policies supporting individual recovery and stimulating societal preparedness. More panel studies in the flood risk domain are needed to gain better insights into the dynamics of individual recovery, risk-reducing behavior, and associated risk and protective factors.     

On predictive distributions and Bayesian networks

In this paper we are interested in discrete prediction problems for a decision-theoretic setting, where the task is to compute the predictive distribution for a finite set of possible alternatives. This question is first addressed in a general Bayesian framework, where we consider a set of probability distributions defined by some parametric model class. Given a prior distribution on the model parameters and a set of sample data, one possible approach for determining a predictive distribution is to fix the parameters to the instantiation with the maximum a posteriori probability. A more accurate predictive distribution can be obtained by computing the evidence (marginal likelihood), i.e., the integral over all the individual parameter instantiations. As an alternative to these two approaches, we demonstrate how to use Rissanen's new definition of stochastic complexity for determining predictive distributions, and show how the evidence predictive distribution with Jeffrey's prior approaches the new stochastic complexity predictive distribution in the limit with increasing amount of sample data. To compare the alternative approaches in practice, each of the predictive distributions discussed is instantiated in the Bayesian network model family case. In particular, to determine Jeffrey's prior for this model family, we show how to compute the (expected) Fisher information matrix for a fixed but arbitrary Bayesian network structure. In the empirical part of the paper the predictive distributions are compared by using the simple tree-structured Naive Bayes model, which is used in the experiments for computational reasons. The experimentation with several public domain classification datasets suggest that the evidence approach produces the most accurate predictions in the log-score sense. The evidence-based methods are also quite robust in the sense that they predict surprisingly well even when only a small fraction of the full training set is used.