Dienstag, 23.04.2024, 12-13 Uhr in W9-109
Julia Dyck, M.Sc.
Universität Bielefeld
Signal detection of adverse drug reactions: The Bayesian Power generalized Weibull Shape Parameter test
After release of a drug on the market, pharmacovigilance monitors the occurrence and changes in known adverse drug reactions (ADRs) as well as detects new ADRs in the population. This is done to keep a drug‘s harm profile updated and can potentially result in adjustments of the prescription labeling or even a recall of the product from the market. In recent years the interest in the use of longitudinal electronic health records for pharmacovigilance increased. Sauzet and Cornelius (2022) provided a test based on the power generalized Weibull (PgW) distribution shape parameters (PgWSP). If both shape parameters of the PgW distribution are equal to one, the distribution reduces to an exponential distribution with constant hazard over time. This is interpreted as no temporal association between a drug and an adverse event. Signal detection can be improved by incorporating existing knowledge about the ADR profile of drugs from the same family or based on expert knowledge about the drug mechanism. Therefore, we propose the development of a Bayesian PgWSP test. The test compares a region of practical equivalence (ROPE) around one reflecting the null hypothesis with the estimated credibility intervals (Krushke, 2015). Depending on the single parameters’ outcome and the chosen combination rule, the tool raises a signal. We performed a simulation study to tune the optimal ROPE and credibility interval for signal detection using a Bayesian PgWSP approach. Samples are generated under varying conditions regarding sample size, prevalence, and proportion of adverse events. Prior assumptions considered are no ADR, ADR at the beginning, middle, or end of the observation period. A range of ROPE and credibility interval types as well as combination rules are considered. The optimal ROPE and credibility interval tuning parameters are determined based on the area under the curve.
Dienstag, 07.05.2024, 12-13 Uhr in W9-109
Prof. Dr. Harry Haupt
Universtät Passau
Fixed-event forecasting systems and efficiency measurement
Fixed-events forecasting systems are regularly used in economic or political analysis. Prominent examples are professional forecasters for key macroeconomic variables or opinion researchers for elections. Despite their widespread use as prima facie forecasts and inputs for economic policy, sentiment analysis or forecasting algorithms, there is considerable uncertainty regarding the definition and measurement of the efficiency of such forecasting systems. The prevailing paradigm defines efficient forecasts by sequential revisions of forecasts that represent martingale differences. Alternative concepts require either knowledge of the forecasters’ dgp or their loss and allow the derivation of criteria analogous to optimal time series forecasts. This paper first examines the theoretical conditions for the weak efficiency concept of forecasting systems for fixed events based on L2-loss and the conditional expectation function. Our main contribution are statistical extensions of existing concepts and their application to German elections between 2000 and 2023.
Dienstag, 21.05.2024, 12-13 Uhr in W9-109
Prof. Dr. Göran Kauermann
Institut für Statistik der Ludwig-Maximilians- Universität München
Statistical Contributions in Conflict Research
The talk introduces to the field of statistical conflict research. We consider armed conflicts in (regions of) Africa with at least one fatality. As data source we make us of the quite reliable data provided by the Peace Research Institut Oslo (PRIO). In the first part of the talk we demonstrate how statical models can be used for predicting future conflicts. We apply a hurdle model, which allows to properly incorporate quantities, relavant from the political science perspective. Secondly, we look at Syria and explore how remote sensing data from satellites can improve prediction accuracy. In particular we include machine learning tools and compare them with statistical approaches. We find the remote sensing can indeed increase the precision of prediction. Finally, we look at the spread of conflicts in time and space, which is often omitted in purely machine learning approaches. The talk ends by posing numerous remaining interesting research questions.
Dienstag, 04.06.2024, 12-13 Uhr in W9-109
Prof. Dr. Dietmar Bauer
Universität Bielefeld
Estimating multinomial probit models using the R package 'Rprobit'
In the talk I present the main functionality of the R package Rprobit currently under development at Bielefeld University. The package can be used in order to estimate (ordered or unordered) multinomial probit models, focusing on the modelling of unobserved heterogeneities in the population under investigation. The talk will touch upon different classes of models relevant for this area. In particular, I will try to convey the interplay between research and coding. Which questions did steer our developments? Which answers prompted directions of further research? As sideremarks I will point out places where master theses played a role in advancing the package. The usage of the package will be demonstrated using a real world data set.
Dienstag, 18.06.2024, 12-13 Uhr in W9-109
Dr. Alfredo Sánchez-Tójar
Universität Bielefeld
Publication bias: what is it and why does it matter?
In this talk, I will (1) introduce the concept of publication bias and discuss different types of publication biases often observed in the literature, (2) present examples of publication bias drawn mostly from the field of ecology and evolution, and (3) briefly discuss how much of a problem publication bias is and how we could study and prevent it.
Dienstag, 02.07.2024, 12-13 Uhr in W9-109
J.-Prof. Dr. Timo Adam
Universität Bielefeld
Statistical modelling of narwhal diving behaviour and responses to sound exposure using stochastic differential equations with state-switching coefficients
Stochastic differential equations (SDEs) are popular tools for uncovering mechanistic relationships underlying time series. In this talk, we propose using SDEs with state-switching coefficients to model narwhal diving behaviour using high-resolution tracking data. For each dive, a non-homogeneous, N-state Markov chain selects which of N possible SDE models determines the depth observations throughout that dive. Using the proposed model, we show that narwhals exhibit two distinct dive types, namely deep, wiggly and shallow, smooth dives. By modelling the transition probabilities of the underlying Markov chain as smooth functions of sound exposure, we further show that, when being exposed to noise, narwhals are less likely to exhibit foraging behaviour.
Dienstag, 16.07.2024, 12-13 Uhr in W9-109
Dr. Tamara Schamberger
Universität Bielefeld
The Sum Score Model: Specifying and Testing Equally Weighted Composites Using Structural Equation Modeling
In principle, structural equation modeling (SEM) is capable of emulating all approaches based on the general linear model. Yet, modeling sum scores in a structural equation model is not straightforward. In this talk, we will present a specification to SEM that allows for directly modeling sum scores and that overcomes existing approaches' limitations in dealing with sum scores in the SEM context. Moreover, the presented approach can mimic the results of existing approaches and provides a means of assessing whether a sum score fully transmits the effects of or on the variables that make up the sum score. In addition, it allows for taking into account random measurement error in the variables that form the sum score.