Psychological network models represent the statistical relationships between discrete psychological variables. These models are theoretically appealing because they can capture systems that display complex behaviors, such as hysteresis or sudden transitions. However, they are also difficult to analyze. Network structures are usually inferred from limited data, and a recent large-scale reanalysis revealed that this uncertainty can be substantial. Ignoring this uncertainty is risky, as it can lead researchers to become overconfident in their conclusions. This workshop introduces a Bayesian approach to psychological network analysis. It treats uncertainty as a central aspect of inference and uses Bayes factor hypothesis testing to evaluate theory-informed questions. For example: Do variables cluster together? Are two variables conditionally dependent? Do networks differ across groups? A key advantage of the Bayesian framework is its ability to provide support for both the presence and absence of effects. This makes it possible to evaluate hypotheses such as conditional independence and across-group equivalence. These are questions that are often central to psychological theory, yet difficult to address with commonly used methods. In addition to covering the theoretical foundations, the workshop will guide participants through the practical application of the Bayesian approach using real data examples. You will learn how to perform Bayes factor tests, interpret the results in the context of network structure, and apply these methods to your own data. By the end of the workshop, you will have both a conceptual understanding and hands-on experience with tools that support transparent and informative network analysis.

Most scientific questions, such as those asked when evaluating policies or exposures, henceforth referred as treatments, are causal in nature, even if they are not specifically framed as such. Causal inference reasoning helps clarify the scientific question, and the assumptions necessary to express it in terms of the observed data. Once this is achieved, the focus shifts to estimation and inference. Estimating causal effects typically requires adjustment for confounding. This is the result of a lack of comparability between subjects due to possibly many factors that are related simultaneously to the outcome and the variable whose effect we aim to estimate. These adjustments can be achieved via parametric modelling. However, such traditional statistical tools are not entirely satisfactory as high-dimensional confounding is difficult to handle and model misspecification is likely. As even minor misspecifications can induce large bias in the treatment effect estimate, the task of learning functional relationships between variables in order to adjust for confounding is critical. Unsurprisingly, machine learning methods are increasingly being used to assist in this task. This is challenging because, while the prediction performance of a given machine learning algorithm can be measured by contrasting observed and predicted outcomes, performance evaluation becomes impossible for treatment effect estimation since the `ground truth’, i.e. the true treatment effect, is unknown. This workshop will start with an introduction to causal diagrams and causal reasoning, and then introduce machine learning-based methods for the evaluation of (causal) treatment effects. I will highlight that bias can be introduced if using standard machine learning methods that are tuned for prediction performance, as opposed to estimation of treatment effects. I will then introduce the framework of “Targeted Learning” and other causal (debiased) machine learning approaches, as a principled solution with optimal statistical properties for the estimation of causal treatment effects. The primary focus will be on learning average total, direct and indirect treatment effects. The workshop will include a number of hands-on sessions in R where participants can experience the problems with naive machine learning and understand how Debiased and Targeted Learning works by implementing it in real-world settings.