Dienstag, 19.04.2016, 12-13 Uhr - Raum: W9-109
Dr. Odile Sauzet
Universität Bielefeld
A mixture Weibull-Exponential model approach for signal detection using longitudinal electronic health records
Signal detection methods are employed to identify previously unknown associations between a drug and an adverse event (AE) of interest. AEs are health events occurring during therapy that may be of unknown cause. Detecting signals for adverse drug reactions is not only important for drugs newly released to the market, but also for established drugs as new safety alerts may still emerge. Detection of new signals can result in changes to prescribing labels or product withdrawal. The increasing availability of longitudinal electronic health records (EHRs) with validated pharmacotherapy recording makes it possible to associate drug prescriptions with health outcomes over time and this offers considerable opportunities for improved pharmcovigilance. In previous work we proposed that detecting a non-constant hazard using the Weibull model could flag a signal for time-dependent ADRs. The Weibull Shape Parameter test (WSP) makes full use of EHR data by detecting a temporal relationship between increased risk and start of treatment and does not require a control group. However this method requires multiple testing which reduces its power. Therefore we propose an alternative method based of a mixture distribution Weibull- Exponential which fits this type of data better and therefore allow for more power. An Adverse event is flagged if the estimated shape parameter of the Weibull part of the mixed distribution is different from one. We describe the statistical method, the algorithm used to obtain the necessary estimates and the possibilities of validation using simulations to compare the specificity and sensitivity of the two methods.
Dienstag, 03.05.2016, 12-13 Uhr - Raum: W9-109
Dr. Nadja Klein
Georg-August-Universität Göttingen
(Multivariate) Structured Additive Distributional Regression
Classical regression models within the exponential family framework such as generalised linear models or generalised additive models focus exclusively on relating the mean of a response variable to covariates but neglect the potential dependence of higher order moments or other features of the response distribution on covariates. As a consequence, the advantage of obtaining covariate effects that are straightforward to estimate and easy to interpret is at least partly offset by the likely misspecification of the model. Structured additive distributional regression provides a generic framework for inference in regression models in which each parameter of a potentially complex response distribution and not only the mean is related to a structured additive predictor. We propose a unified Bayesian approach for multivariate structured additive distributional regression comprising a huge class of continuous, discrete and latent multivariate response distributions, where each parameter of these potentially complex distributions is modelled by a structured additive predictor. The latter is an additive composition of different types of covariate effects e.g. nonlinear effects of continuous covariates, random effects, spatial effects, or interaction effects. As a flexible approach for constructing tailor-made multivariate response distributions, we consider copula-based regression models since they enable the separation of the marginal response distributions and the dependence structure summarised in a specific copula model. Inference is realised by a generic, computationally efficient Markov chain Monte Carlo algorithm based on iteratively weighted least squares approximations and with multivariate Gaussian priors to enforce specific properties of functional effects. The approach is illustrated along a joint model for different risk factors measuring childhood malnutrition in India.
Dienstag, 17.05.2016, 12-13 Uhr - Raum: W9-109
Prof. Dr. Göran Kauermann
Institut für Statistik der Ludwig-Maximilians-Universität München
Exponential Random Graph Models with Stable (Non-Parametric) Statistics
Exponential Random Graph Models (ERGM) are a common tool for network data analyses. The ERGM describes the distribution of a network graph with an exponential family distribution, where the statistics are counts of edges, k-stars or triangles, for instance. Even though this allows for simple and meaningful interpretations of network data the models are notoriously unstable if the network is large, i.e. if there are several hundred nodes in the network. To overcome this problem a number of modifications have been pursued, including curved exponential models and geometrically down-weighted statistics, which come however for the price of reduced interpretability. We propose to make use of simple statistics like k-stars or triangles, which allow for intuitive and meaningful interpretations. But instead of using linear forms of the statistics we use non-linear versions, where the latter are either fitted by non-parametric, monotonic splines or by pre-specified curved functions. The approach provides simple interpretations of network data by maintaining stability of the model.
Dienstag, 31.05.2016, 12-13 Uhr - Raum: W9-109
Holger Steinmetz
Universität Paderborn
MASEM: The integration of meta-analysis and structural equation modeling
Meta-analyses and structural equation modeling (SEM) are two powerful methods which, however, have developed separated from each other, spread across different, fields and are seldom used in combination. This may partly results from their diverging goals: Meta-analyses are a potent method to aggregate and synthesize empirical results and, hence, aim at integrating the existing literature. SEM, in contrast, is applied to primary data and aims at specifying and testing causal structures. In recent years, however, attempts have been made to develop meta-analytical structural equation models (MASEM) as the “best of both worlds” – hence to test causal structures with meta-analytically derived, aggregated data. The presentation introduces in the main principles of both meta-analysis and SEM, describes the strengths and shortcomings of both methods, and illustrates the process of conducting a MASEM analysis.
Dienstag, 14.06.2016, 12-13 Uhr - Raum: W9-109
Peter Pütz
Georg-August-Universität Göttingen
A Penalized Spline Approach For Fixed Effects Panel Data Models
Estimating nonlinear effects of continuous covariates by penalized splines is well established for regressions with cross-sectional data as well as for panel data regressions with random effects. Penalized splines are particularly advantageous since they enable both the estimation of unknown nonlinear covariate effects and inferential statements about these effects. The latter are based, for example, on simultaneous confidence bands that provide a simultaneous uncertainty assessment for the whole estimated functions. In this paper, we consider fixed effects panel data models instead of random effects specifications and develop a first-difference approach for the inclusion of penalized splines in this case. We take the resulting dependence structure into account and adapt the construction of simultaneous confidence bands accordingly. In addition, the penalized spline estimates as well as the confidence bands are also made available for derivatives of the estimated effects which are of considerable interest in many application areas. As an empirical illustration, we analyze the dynamics of life satisfaction over the life span based on data from the German Socio-Economic Panel (SOEP). An open source software implementation of our methods is available in the R package pamfe.
Dienstag, 28.06.2016, 12-13 Uhr - Raum: W9-109
Jens Möller
Bielefeld Marketing GmbH
Bielefeld auf dem Weg zur starken Stadtmarke
Ziel des neuen Stadtmarkenprozesses ist es, starke Bielefeld-Themen zu erkennen, zu priorisieren und gemeinsam zu kommunizieren. Für die Markenanalyse ergibt sich damit die zentrale Fragestellung: "Was macht Bielefeld für seine Bewohner und Gäste besonders attraktiv?" Zur Vorbereitung der Online-Befragung wurden insgesamt sechs verschiedene Workshops mit insgesamt 70 Bürgern, Experten und Wissensträgern aus Bielefeld und der Region Ostwestfalen durchgeführt. Im zweiten Teil der Markenanalyse spielte die persönliche Meinung und Wahrnehmung der Menschen aus Bielefeld, der Region Ostwestfalen-Lippe und darüber hinaus die entscheidende Rolle. 94 konkrete Bielefelder Gegebenheiten wurden von den Befragten hinsichtlich ihrer Attraktivität bewertet. Die Ergebnisse der Befragung fließen als wichtige Grundlage in die Entwicklung der neuen Vermarktungsstrategie für Bielefeld ein.
Dienstag, 12.07.2016, 12-13 Uhr - Raum: W9-109
Aktuelle Forschungsbereiche im ZeSt