Methods Center

Workshop 1 - Vasishth / Nicenboim

Introduction to Bayesian Modeling using Stan

Shravan Vasishth (with Bruno Nicenboim), University of Potsdam

 

In this one-day workshop, we will give a comprehensive introduction to using Stan for Bayesian data analysis and Bayesian modeling.

By the end of the course, participants should be able to:

Understand the syntax of Stan and the high-level ideas behind MCMC and HMC.
Fit standard models (such as linear models) in Stan.
Understand how hierarchical modeling works, and be able to fit complex hierarchical models in different settings.
Carry out sensitivity analyses to investigate how posteriors change as a result of prior specification.
Visualize and interpret different models.
Carry out posterior predictive checks and cross-validation for model evaluation.
We will provide lecture notes and suggested readings for further study. We assume that everyone has a laptop with them and has the R package rstan installed within R.

This one-day workshop will involve lectures interspersed with short exercises to be done in class. In order to consolidate understanding, we will assign a project that participants can carry out (this is optional). Students have the option to submit it to the instructor a week later and get feedback.

More information on the workshop are provided here: http://www.ling.uni-potsdam.de/~vasishth/courses/IntroStanFGME2017.html

Workshop 2 - Rosseel / Mayer

Multilevel Structural Equation Modeling with lavaan

Yves Rosseel and Axel Mayer

 

The aim of this workshop is to provide an introduction to the multilevel structural equation modeling (SEM) framework with lavaan. We focus on the application of this framework to analyze multilevel data (for example: student scores, where students are nested in schools). First, we will discuss the relationship between classic (single-level) regression, multilevel regression, and SEM. We will do this both from a theoretical point of view as well as from a software point of view. We will show how and under which conditions (classic, non-multilevel) SEM software can produce identical results as dedicated multilevel (or mixed modeling) software. Second, we will demonstrate how lavaan can be used to analyze multilevel data. We will start from a regression perspective, and gradually proceed from a simple regression analysis, to a two-level regression analysis, towards more complicated models, exploiting the full power of the multilevel SEM framework. Third, we will take a latent-variable (CFA) perspective, and give various examples of multilevel CFA, and multilevel SEM involving latent variables. Fourth, we will introduce moderation and mediation within the multilevel SEM framework and provide some examples. Along the way, we will discuss many practical issues including the role of centering, the treatment of missing and/or non-normal data, and how to deal with categorical data.

lavaan-Link: http://lavaan.ugent.be/

Workshop 3 - Muma

Robust Estimation with Applications in Psychophysiological and Biomedical Signal Processing

Michael Muma

 

Statistical signal processing is a powerful area of research that has been successfully applied to generations of research problems in order to extract useful information from empirical data. An effective way to incorporate knowledge about the application at hand is to use parametric models.  When applying parametric methods to real-world problems, it often happens that the observations do not exactly follow the assumptions that were made to model the problem. In these cases, the nominally excellent performance can drastically degrade.

This seminar introduces the concept of robust estimation in a way that it is accessible to researchers in the area of psychology. Robust statistics formalize the theory of approximate parametric models. On the one hand, they are able to leverage upon a parametric model, but on the other hand, they do not depend critically on the exact fulfillment of the model assumptions.

The basic concepts and foundations of robust statistical theory are introduced by considering the robust estimation of the location and scale parameters of a univariate random variable. In particular, measures such as the influence function, the breakdown point and the trade-off between robustness and statistical efficiency are discussed.

After laying the foundations using simple models the concepts of robust statistics are extended to more challenging situations. Parameter estimation in linear regression is considered because of its importance of modeling many practical problems. Next, the theory of robust estimation of the multivariate location and scatter (covariance) matrix is introduced and an application of robust regularized discriminant analysis for emotion classification is provided in detail. Finally, correlated data streams, which are commonly measured, e.g., in psychophysiology are treated. A focus will lie on robust parameter estimation for ARMA models as well as methods of robust filtering and outlier cleaning.

During the entire tutorial, ample real-life applications from engineering, bio-medicine and psychophysiology will be given along with information to publicly available implementations of the considered algorithms.

Workshop 4 - Stemmler

Personen-orientierte statistische Methoden oder die Analyse von Kontingenztafeln mit R: Log-Lineare Modelle und Konfigurationsfrequenzanalyse (KFA)

Mark Stemmler

 

Während die Variablen-orientierte Psychologie Mittelwerte oder Kovarianzen von Skalenwerten psychologischer Konstrukte untersucht, beschäftigt sich der Personen-orientierte Ansatz mit Merkmalskombinationen, die eine Person auf sich vereinigt und mit deren Auftretenshäufigkeiten. Merkmalskombinationen oder Konfigurationen, die signifikant häufiger auftreten als unter der Nullhypothese vorausgesagt werden als Typen bezeichnet. Konfigurationen, die signifikant seltener auftreten als unter der Nullhypothese postuliert werden als Antitypen bezeichnet (Krauth & Lienert 1973). Die vorgestellten Verfahren eignen sich auch für kleinere Stichproben oder für den Fall, dass die statistischen Voraussetzungen für Multivariate Verfahren wie Intervalldaten, Normalverteilung oder homogene Fehlervarianzen nicht erfüllt sind.

Neben einer Einführung in die open-source Software R, wird das R-Paket confreq  vorgestellt. Es ermöglicht nicht nur die Analyse von Konfigurationen mit der KFA, sondern auch die Analyse von log-linearen Modellen sowie die zwei-Stichproben-KFA, die als Pendant zum t-Test für Kontingenztafeln angesehen werden kann. In der neuesten confreq Version können auch Kovariaten verwendet werden. Alle Methoden werden anhand von kleinen Zahlenbeispielen demonstriert und am eigenen Laptop geübt.

Der Personen-orientierte Ansatz wird hier nicht als Entweder-oder-Alternative zum Variablen-orientierten Ansatz präsentiert werden, sondern er soll die Denk- und Auswertungsmöglichkeiten in der psychologischen Forschung erweitern.

Voraussetzungen für Teilnahme: Verlangt werden nur Grundkenntnisse in multivariaten Verfahren; der Kurs bietet gleichzeitig eine Einführung in die Software R!

Literatur:

Stemmler, M. (2014). Person-centered  methods: Configural Frequency Analysis (CFA) and Other Methods for the Analysis of Contingency Tables. Series: Springer Briefs in Statistics. New York: Springer Publishing Company.

Stemmler, M. & Heine, J.-H. (2016). Using Configural Frequency Analysis as a Person-Centered Analytic Approach with Categorical Data. International Journal of Behavioural Development. DOI: 10.1177/0165025416647524.

Zielgruppe:

Die TeilnehmerInnen sollten Kenntnisse der Inferenzstatistik sowie der multivariaten Analyseverfahren mitbringen. Gearbeitet wird am eigenen Laptop, auf dem bereits die R-Version 3.1.0 (oder neuer) sowie ein R-Editor (am besten R-Studio) installiert sind. Einige Datenbeispiele werden auch mit Hilfe von SPSS analysiert.

Workshop 5 - Regenwetter

QTEST: Quantitative Testing of Theories of Binary Choice

Michel Regenwetter

 

This half-day workshop is aimed at graduate students, postdocs and faculty interested in learning about distribution-free models of binary choice: How to build them and how to test them. The format will combine a lecture with hand-on exercises. Attendees should bring a laptop with the QTEST software installed. While the workshop does not require prior reading, attendees will benefit from reading any or all of the papers (reprints at internal.psychology.illinois.edu/reprints/index.php?site_id=38) listed below.

Key papers:

Regenwetter, M., Davis-Stober, C.P., Lim, S.H., Cha, Y.-C., Guo, Y., Messner, W., Popova, A., and Zwilling, C. (2014). “QTEST: Quantitative Testing of Theories of Binary Choice.” Decision, 1,1, 2-34.

Regenwetter, M. Dana, J. & Davis-Stober, C. (2011). "Transitivity of Preferences." Psychological Review, 118, 684-688.

Also relevant:

Brown, N. R., Davis-Stober, C.P., and Regenwetter, M. (2015). “Commentary: “Neural signatures of intransitive preferences” ” Frontiers in Human Neuroscience.

Davis-Stober, C., Park, S., Brown, N. and Regenwetter, M. (2016). “Reported violations of rationality may be aggregation artifacts.” Proceedings of the National Academy of Sciences of the United States of America.

Guo, Y. and Regenwetter, M. (2014).“Quantitative Tests of the Perceived Relative Argument Model: Comment on Loomes (2010).” Psychological Review, 121, 696-705.

Regenwetter, M., Cavagnaro, D., Popova, A., Guo, Y., Zwilling, C., Lim, S.H., Stevens, J.R. (in press). “Heterogeneity and Parsimony in Intertemporal Choice.” Decision.

Regenwetter, M. and Davis-Stober, C.P. (2012). "Behavioral Variability of Choices versus Structural Inconsistency of Preferences." Psychological Review, 119, 408-416.

Regenwetter, M. and Robinson, M. (in press). “The construct-behavior gap in behavioral decision research: A challenge beyond replicability.” Psychological Review.

Workshop 6 - Hufer

Introduction into behavior genetic analyses using data from TwinLife (A genetically informative study on the development of social inequality)

Anke Hufer

 

TwinLife is a genetically informative study on the development of social inequality, using a cross sequential-design. During the first wave of data collection, data concerning numerous different core areas of social inequality has been collected from 4.097 families (including overall 19.163 individuals).

In a first part of this workshop, we will give you an overview of the TwinLife study in general – design, constructs, and focuses of content, the open source dataset. On the basis of selected examples, we will give an introduction of behavior genetic analyses to get an idea of how you could use it for your own research questions. In the second part of the workshop, participants are invited to participate in practical exercises. Here you will learn how to implement this kind of structural equation models on genetically informative data. It’s also intended to get familiar with the data, which is available for researchers for free.

Especially for the practical session, our participants are requested to bring their own laptop. A basic understanding of structural equation models is very helpful.

Prior to the workshop, participants will be contacted to make sure that all technical requirements are clearly defined to enable a smooth running.

Maximum number of participants: 20

Please remark that the workshop can be cancelled if there are not enough participants. In any cases, you will be informed in time.

The participation at this workshop is free of charge. For registration or questions, please contact Anke Hufer (anke.huferspam prevention@uni-bielefeld.de). The deadline for the registration is the 20. August.
The registration will not be via the ConfTool.