Workshop Teaching Record

I started doing methods workshop teaching still during my PhD. Communicating research methods for the next generation of scholars quickly became a passion of mine. Today, I am one of the three people responsible for the academic content of the European Consortium for Political Research Summer and Winter School in Methods and Techniques, one of the largest and definitely the most diverse multi-method social science methods training schemes in the world. Through the journey from short workshops during my PhD to heading one of the best methods schools in the social sciences (…yes, I am biased but also proud of what we do), I have developed and taught multiple courses myself. Here is the list as it stands today:

Multilevel Modeling – To teach multilevel modeling I start with the basics of a fixed effects regression model (regression correcting heterogeneity with dummy variables) and expand into random effects. I do this with caution pointing out the problems of small samples, non-random samples, our inability to control for most relevant factors on the country level, the most common problems associated with using multilevel modeling for cross-national surveys as done so often in political science and sociology. I emphasize the power of centering in separating level 1 and level 2 effects, the use of multilevel models for panel data and extend into three-level and cross-classified models. (Workshop is done in R. I provide scripts for Stata, SPSS and Mplus in addition.) Teaching experience: ECPR Winter School 2013, 2014, 2015, 2016, 2017, 2018, 2019, University of Zagreb (2008), and PhD courses at CEU (2016, 2019).

Structural Equation Modeling – Today’s teaching texts for structural equation modeling have gotten unnecessary long in later editions. For this reason Jochen Mayerl (ECPR Summer School instructor for SEM) and I set out to develop a short, to the point SEM text (forthcoming) for the prestigious SAGE Quantitative Applications in Social Sciences (little green) book series. Examples used in the text revolve at the intersection of political science, psychology and sociology focusing on the impact of authoritarianism on prejudice. We use examples in Lavaan (for R) and/or in Mplus. The text starts with the assumption that the reader knows regression models and the basics of factor analysis (or other latent variable models). It explores debates of the field on model specification from a theoretical point of view. Teaching experience (beyond writing the book): workshop at Tallinn University (2018), VSE Prague (2018), and PhD course at CEU (2017).

Advanced Structural Equation Modeling – I developed a course for social scientists who have a solid foundation in the basics of structural equation modeling and would like to approach some of the advanced topics. These include the use of exploratory structural equation models, the use of link functions within structural equation models to deal with various distributions of observed variables (including an exploration of the connection between IRT models and CFA with a logic or probit link). An in depth analysis of missing data in SEM. We get into longitudinal models exploring AR1 structures, time-invariant random measurement error corrections and latent growth curves. Finally, we explore how the logic of latent growth models extends into the world of multilevel structural equation models. This course is quite modular and could include other topics such as the exploration of invariance (including the alignment method), mixture extensions of SEM, the novel behavior genetic models for twin and family designs and etc. Teaching experience only offered once at the 2017 ECPR Summer School. (This course requires the full version of Mplus.)

Multilevel Structural Equation Modeling – With two CEU PhD students at the time, we developed a book on multilevel structural equation modeling. To date, texts covering this topic existed only as scattered chapters in multilevel modeling or structural equation modeling textbooks. This is the first, and for now the only devoted entirely to the subject in the prestigious SAGE Quantitative Applications in Social Sciences (little green) book series. Examples span the fields of cross-cultural social science surveys, education research data and organizational behavior research. (This workshop requires Mplus, but all examples brought run on the free demo version.) Teaching experience: ECPR summer school (2016 and 2019) and Workshop at the University of Bamberg (2017).

Experimental Political Science – This course was developed to be more a substantive political science course exploring examples of good and less good political science research taking an experimental approach. Topics included lab, survey, field and natural experiments and potentially also quasi-experimental designs for causal inference using observational data. (Statistical background is not necessarily needed, though basics of statistical inference is extremely useful. No software necessary.) Teaching experience: MA Class at CEU 2016, 2017, 2018. I would be ready to offer a more methods oriented Experimental Design course as well focused less on examples of experimental research in political science and more on a textbook like the Morton and Williams Experimental Political Science and the Study of Causality: From Nature to the Lab.

Advanced Topics in Regression Models – Under several titles I have conducted a class aimed at more careful thinking about regression. This could be a third class in regression where we dive deep in the nuanced topics. The course starts with a review of the assumptions of regression models and then dives deep into the potential assumption violations. The topics covered can include any mixture of of the following topics: issues of measurement and operationalization with a focus on measurement error; modeling of observed and unobserved heterogeneity with fixed and random effects, and mixture models; bootstrapping; collinearity and its remedies; sampling weights and the use of weights to correct for unit non-response; multiple imputation for item non-response; relaxing linearity through splines and nonparametric preprocessing; alternative model specifications and model averaging. (Workshop is done in R.) Teaching Experience: ECPR Summer School 2012, 2014, 2015, 2016

Missing Data – Missing data was one of the first things that fascinated me to pursue a focus on methods. How do you analyze data you don’t have. (Ecological inference was the other, which is also a missing data problem in a way.) Missing data is one of these issues that people try to ignore, and it is a really bad idea. If you have missing data, you are doing something with it. I strongly believe that the best way to deal with missing data is not to let the software defaults kick in. Rather you want to understand the biases and inefficiencies associated missing data, the assumptions missing data corrections make. You need to learn to think theoretically about missing data. This workshop accomplishes exactly this. We learn how to think about missing data. Review the classic approaches to missing data correction with an emphasis on what is wrong with them. We look at modern techniques, discuss the relative advantages and disadvantages of these. Finally, we practice in sets designed to highlight the biases and inefficiencies associated with various missing data assumptions and procedures to correct them. (Workshop is done in R. Could add an Mplus component.) Teaching Experience: ECPR Summer and Winter School (2012, 2017), Tartu University (2013), PhD course at CEU (2009, 2014), University of Zagreb (2007), University of Nebraska-Lincoln (2007).

Causal Inference – Continuing a long lineage of borrowing our intellectual foundations from economics, there is an increased emphasis in political science on causal claims or the elimination endogeneity or reverse causal effects in the rest of relationships. This is not easy to. The gold standard for the test of causal relationships is still randomized controlled trials. But what if the collection of experimental data is simply not possible. Building on Rubin’s causal model, we can generalize the experimental logic and still make causal claims using observational data. These, so-called quasi-experimental approaches can include granger causal models, encouragement designs and more complex instrumental variable models (very popular in econometrics nowadays), regression discontinuity designs, difference-in-differences models and matching. In a longer course format we can review synthetic controls for single case longitudinal causal inference and also how structural equation models can produce equivalent results to the above mentioned approaches with greater flexibility. (Workshop done in R.) Teaching Experience: Tartu University (2013), European Policy Center in Belgrade, Serbia (2017), CEU (2009, 2014).

Program Evaluation for Practitioners (without statistics) – I have developed a course that covers the logic of counterfactual program evaluation (basically the techniques designed for causal inference in the social sciences) without any statistics background. This workshop is useful for practitioners who need to design state of the art evaluation programs, at the same time allows them to leave the number crunching with trained analysts. Topics, once again cover randomized controlled trials, encouragement designs, regression discontinuity designs, difference-in-differences models, matching, and synthetic controls. The workshop is built on the second edition of the Impact Evaluation in Practice book made available for free by the World Bank in English, Spanish and French. Teaching Experience: University of Granada, Faculty of Education (2019); Think for Europe Network organized by European Policy Center in Belgrade, Serbia (2019).

I also have a course developed for absolute beginners of Inferential Statistics (who are typically afraid of the subject) using the Gravetter and Wallnau Essentials… textbook (taught yearly at CEU and at ECPR summer school 2017, 2018, 2019); an Introductory Programming and Analysis in R course to complement it (taught yearly at CEU and at Tallinn University 2018); and an Introductory Regression course also with R applications (taught yearly at CEU, ECPR Summer and Winter School 2018, 2019, Tallinn University 2018). I teach these every year at CEU. Currently I am working on developing an Introduction to Quantitative Text Analysis for Political Scientists course using Quanteda for R.

Not methods but, in the past, I have also been invited to teach Behavior Genetics of Political Science (Lauder School of Government, Diplomacy and Strategy, Interdisciplinary Center, Herzliya, Israel. 2013).