EAM 2018 - Pre-conference workshops

There were pre-conference workshops on July 24, 2018.

Intensive Longitudinal Data Analysis / DSEM

Utrecht University, the Netherlands

In this workshop participants will learn the basics of Dynamic (Multilevel) analysis of intensive longitudinal data, which is the basis for Dynamic SEM (DSEM). These models are also what is used in the background for creating dynamic network models in Psychology.

We start by covering what intensive longitudinal data is, and why these data and the single subject and multilevel dynamic models for analyzing them are of interest. The single subject analyses can be used to model the dynamics within a particular individual over time, such as the reciprocal effects of positive affect and worrying may have on each other. The multilevel extensions of these models can be used to model the individual within-person dynamics at level 1, and the individual differences in these dynamics at level 2. We will discuss the basics of univariate and multivariate single subject autoregressive models, and the multilevel extensions of this model. We discuss various software options (in and outside of R) and their benefits and limitations, and give a live demonstration of the models in the DSEM module of Mplus v8. This includes a hands on part in R (bring a laptop with R; or a laptop with Mplus v8 if the participant has a license).

After the basics we move on to discuss further extensions of the models, such as predicting individual differences in dynamic effects with other level 2 variables, using the individual differences as predictors of level 2 variables, and accounting for measurement error with latent variables. We end with a discussion of potential caveats and pitfalls, recent developments, and what developments to expect for these models and data in the future.

Multilevel Structural Equation Modeling with Lavaan

Yves Rosseel

Ghent University, Belgium

The aim of this half-day workshop is to provide an introduction to multilevel structural equation modeling (SEM). After a brief history, and an overview of the different frameworks, we will focus on the two-level within/between approach that is most commonly used in the applied literature. Special focus will be given to the status/meaning of latent variables in a multilevel setting, and the distinction between observed and latent covariates. Several examples will be discussed, including the setup in lavaan. Finally, we will discuss several alternative approaches to multilevel SEM, and explain when they should be used.

Understanding SEM: Where do all the numbers come from?

Yves Rosseel

Ghent University, Belgium

This half-day workshop will offer a glimpse behind the scenes of SEM software. In the first part, it will be demonstrated how the model parameters in a SEM are estimated. We will set up a small function (in R) that takes a parameter vector as input, and computes the value of a suitable discrepancy function. We will then exploit the built-in optimizers of R (optim, nlminb) to find the best fitting parameters. In the second part, we will discuss the concept of information matrices, and we will explain how both standard and robust standard errors can be computed. Finally, in the third part, we will show how the so-called robust (Satorra-Bentler, Yuan-Bentler) test statistics are computed. All of this will be demonstrated with base R. This workshop aims to deepen your understanding of how Structural Equation Modeling works.

Hypothesis evaluation using the Bayes factor

Herbert Hoijtink

Utrecht University, the Netherlands

This workshop will introduce the participants to null hypothesis significance testing and its role in the replication crisis. Subsequently, an alternative, hypothesis evaluation using the Bayes factor will be introduced. It will be elaborated what the Bayes factor is, how it can be applied and should be interpreted. There will be attention for Bayesian updating (an alternative for power analysis), Bayesian (conditional) error probabilities, limitations of the approach, and software with which the Bayes factor can be computed.

Theory and analysis of conditional and average causal effects

Rolf Steyer

University of Jena, Germany

This short course is an introduction to the stochastic theory of causality, which is a generalization of the theory of causal effects in the tradition of J. Neyman and D. B. Rubin. In the course I will present the stochastic theory of causal effects and show how to use EffectLiteR for the analysis of conditional and average total effects.

Contents

Motivation: Simpson's paradox, non-orthogonal ANOVA
The scope of the theory: random experiments
The mathematical structure of causal models: causality space
True outcome variables, average and conditional causal effects
Prima facie effects
Sufficient conditions for unbiasedness
The role of randomization and other design techniques and strategies of data analysis
Estimating and testing average and conditional total effects via structural equation modeling (Applications using EffectLiteR) in some empirical examples