Logo Utrecht University

S4 conference

Post-conference workshops

We will close the conference on June 11 with a day filled with workshops. Each three-hour workshop
discusses one specific solution is great detail including some hands-on exercises. Registration is for one
specific workshop and always includes lunch (after or before your workshop) and you can combine
a morning workshop with one in the afternoon.

Morning Workshops

Hypotheses Evaluation using the Bayes Factor
In the last decade there has been increasing attention for informative hypotheses (e.g., m1 > m2 > m3, where the m’s denote group means) in addition to the traditional null (e.g., m1 = m2 = m3) and alternative hypotheses. The reason for this is that the expectations of researchers are often better represented by informative hypotheses. Furthermore, there is a still increasing attention for the use of Bayes factors instead of p-values to evaluate hypotheses. In this workshop participants will be familiarized with both informative (and classical) hypotheses and their evaluation using the Bayes factor. It will also be shown how hypotheses evaluation using the Bayes factor can easily be executed using the bain R package (https://informative-hypotheses.sites.uu.nl/software/bain/). The workshop is tailored to social and behavioral scientists, that is, concepts will be introduced, explained and exemplified, but formulas will not play a role in this workshop. If you like a preview of the content of the workshop, you should read

Hoijtink, H., Mulder, J., van Lissa, C., and Gu, X. (2018). A tutorial on testing hypotheses using the Bayes factor. Psychological Methods. DOI: 10.1037/met0000201

which can be downloaded from the bain website.
Integrating information beyond data: How to obtain informative priors
Bayesian estimation with (weakly) informative priors can be a solution for small sample size situations. Naïve use of Bayesian statistics, in contrast, can actually cause harm. In this workshop we show where and how to find additional information to specify (weakly) informative priors for your models. Amongst others, we discuss natural parameter constraints and the elicitation of expert knowledge.
Bayesian analysis of multiple binary outcomes
This workshop provides a practical introduction to the Bayesian analysis of multiple binary outcomes. Participants will be guided through the steps required to compare data from two groups, using freely available online software. In addition, several design-related topics will be covered, such as the specification of a suitable prior distribution and the selection of an appropriate sample size.
Using Multilevel Modeling to Meta-Analyze Single-Case Experimental Design Studies’ Results
This workshop introduces the use of multilevel models for meta-analyzing single-subject experimental design (SSED) studies. The course will begin with an overview of multilevel modeling and how the model can be used for SSED meta-analysis of unstandardized data. Next, an overview of how to standardize raw data or effect sizes will be given, and the meta-analysis of these standardized data will be demonstrated. Multiple, real-world examples will be provided as well as opportunities for guided practice for participants. Copies of R-code for every analysis will be provided. Participants need only be familiar with multiple regression analysis and the principles of statistical inference.
Parceling in small sample sizes
In this workshop we will discuss the benefits of parceling in small sample sizes. Parceling addresses several issues stemming from sources of variance commonly found in item level data. We will talk about how parcels can increase power in small sample sizes, avoid convergence issues by reducing model complexity, produce more reasonable type I error rates and avoid the over-parameterization (and therefore mis-specified models) seen in small sample size studies. We will review examples of the above-mentioned topics in this workshop.

Afternoon workshops

Rep Lab
In this workshop we will test whether replication attempts actually replicate the original findings. We will discuss what the replication question actually is, and apply the appropriate replication evaluation tool.
Shrinkage priors to fit big models on small samples
During this workshop we will discuss shrinkage priors. Shrinkage priors are especially useful in models that are large compared to the amount of data available, for example regression models with more predictors than observations. You will learn when and how to use shrinkage priors and get some hands-on experience using the software program Stan.
Dangers of Default Priors
A workshop on the impact of default priors when Bayesian Structural Equation Modeling (BSEM) is used with small samples; how to recognize misbehaving priors; and recommendations on what (not) to do when BSEM is used in small sample contexts.
restriktor: beyond classical null-hypothesis testing
Confirmatory approaches, such as the GORIC (an AIC-like criterion), have proven to be more `powerful' (i.e., render a higher probability of choosing the best hypothesis) than exploratory approaches such as the AIC. In this workshop, we will demonstrate that if you specify your theory and competing theory wisely, you can increase your power substantially. This is extremely helpful in case of small-sample data sets.

Workshop will be given by Rebecca Kuiper.
Dealing with heterogeneity in meta-analysis
This workshop provides a practical introduction to the Bayesian analysis of multiple binary outcomes. Participants will be guided through the steps required to compare data from two groups, using freely available online software. In addition, several design-related topics will be covered, such as the specification of a suitable prior distribution and the selection of an appropriate sample size.