Your subscription could not be saved. Please try again.

Your subscription to the ECPR Methods School offers and updates newsletter has been successful.

Discover ECPR's Latest Methods Course Offerings

* Provide your email address to subscribe. For e.g abc@xyz.com

I agree to receive your newsletters and accept the data privacy statement.

You may unsubscribe at any time using the link in our newsletter.

We use Brevo as our email marketing platform. By clicking below to submit this form, you acknowledge that the information you provided will be transferred to Brevo for processing in accordance with their terms of use.

Multivariate Statistical Techniques for Comparing Countries

Member rate £492.50
Non-Member rate £985.00

Save £45 Loyalty discount applied automatically*
Save 5% on each additional course booked

*If you attended our Methods School in the July/August 2023 or February 2024.

Course Dates and Times

Monday 15 ꟷ Friday 19 March 2021
2 hours of live teaching per day
09:30-12:30 CET

Bruno Cautrès

bruno.cautres@gmail.com

Institution: Sciences Po Paris

This course provides a highly interactive online teaching and learning environment, using state of the art online pedagogical tools. It is designed for a demanding audience (researchers, professional analysts, advanced students) and capped at a maximum of 16 participants so that the teaching team (the Instructor plus one highly qualified Teaching Assistant) can cater to the specific needs of each individual.

Purpose of the course

The course offers an introduction to the main statistical techniques used to analyse cross-national comparative surveys data.

Its main goal is to teach you how different statistical methods treat the ‘country effect’: how statistical models (linear regression, logit models, loglinear models, multilevel regression models), scaling techniques (from simple methods to complex factorial techniques) or data reduction methods (factor and PCA analysis) test for the ‘invariance’ of the relationship between variables across countries.

The course has two key points:

it offers a reasonable level of formalisation, as much as needed to understand the methods
it makes links between the different methods and the learning of complementarities between methods.

ECTS Credits

3 credits Engage fully with class activities
4 credits Complete a post-class assignment

Instructor Bio

Bruno Cautrès is attached to CEVIPOF – Centre de recherches politiques de Sciences Po (Paris), at the Fondation Nationale des Sciences Politiques in Paris.

He is a senior CNRS research fellow with interests in voting behaviour, political attitudes and behaviours, comparative survey research and quantitative techniques.

Bruno is involved in a variety of projects, including the European Social Survey, European Values Studies, International Social Survey Programme and European elections studies; and he participates in the development of elections studies in France. His current research programme concerns political trust and attitudes to democracy in France.

@BCautres

Key topics covered

We start on Monday with a review of the substantial and methodological problems the multivariate statistical analysis tries to solve for comparing countries, particularly what is a ‘country effect’ and how this translates in statistical terms. The first day’s activities will also show very simple recalls using descriptive and bivariate techniques applied to ‘country effects’ problems.

Then we will work to understand and two main types of multivariate techniques, still applied to the same question of the ‘country effects’:

Tuesday and Wednesday ꟷ statistical modelling techniques such as linear regression, logit and loglinear models
Thursday and Friday ꟷ data reduction techniques such as principal components and scaling techniques.

Finally, on the Friday, we will also extend the perspective to multilevel analysis applied to cross-national data analysis. Examples will be given and replicated through R programmes.

How the course will work online

The course is designed to exploit the interactive capabilities of online technology, combining annotated readings, short pre-recorded lectures, and live group work. It combines pre-class activities and live/online interaction. You will be given annotated and interactive readings which we will discuss ahead of the course. You should have completed the readings and watched the pre-recorded lectures on the key topics. The pre-recorded lectures take two forms: a methodological presentation and an illustration, through a published text or a data analysis, explained in plain English. The main tables or graphics will serve as key elements to understand the substantive issues of the methods.

There will be ten hours, or two hours each day, of live, in-class teaching. During these two hours we will review the main methodological issues and concepts related to the pre-recorded presentations and annotated readings.

We will set up an online Slack community to discuss course-related matters. The instructor will provide R scripts for running analysis and you will develop and complete them as your project work. During office hours, you’ll also have the chance to sign up for a quick one-to-one consultation with the Instructor or TA.

The course will consist of five sessions, organised in three main topics

Topic 1

The substantial and main methodological issues of comparing countries through statistical techniques

What does it mean to test for ‘country effects’; what are they?
How does this translate into statistical terminology and methods?
What are the main notions, concepts and vocabulary used in this research field?

Topic 2

How the reasonings of statistical models do this ‘country effects’ analysis, through classical techniques like linear regression models, logit and loglinear models. We will pay attention to the fundamental problem of comparing regression estimates from one country to another one and the sometimes tricky issues of it.

Topic 3

How data reduction techniques are used, and can be used, in cross-national analysis, through classical techniques like PCA. We will pay attention to another fundamental problem of comparative analysis: can we use the same instruments (survey questionnaire items) across countries and what does it mean to run scaling analysis across it? Friday’s session will provide some short but important extensions to multilevel analysis and structural equations models that will prepare you to attend other courses more specialised in these techniques.

You must be familiar with with basic statistical inference and regression models. If you are not, take the course Introduction to Inferential Statistics before you sign up for this one.

Day	Topic	Details
1	Describing and comparing countries with descriptive statistics, simple way to test for 'country effects': means, proportions; correlations and odds-ratios	Lecture – 90 minutes Basic recall in descriptive statistics and hypothesis testing with a special emphasis on comparing countries. The general framework of the course: to test for country effect or 'invariance'. Lab – 90 minutes Hypothesis testing and testing for country effects with survey data: comparing means, proportions, correlations and odds ratios across countries.
2	The fundamentals of regression analysis	Lecture – 90 minutes Basics of regression analysis: how it works, ANOVA, estimates and average / marginal effects, goodness of fit, OLS assumptions. Lab – 90 minutes Running a regression analysis based on crossnational survey data; weighting or not weighting? Looking at residuals.
3	Controlling for 'country effects in multiple regression analysis	Lecture – 90 minutes The regression framework for crossnational survey analysis: testing for 'country effects', comparing regression estimates from one country to another, avoiding common mistakes in comparing estimates, standardised estimates. Lab – 90 minutes Running a multiple regression analysis on crossnational surveys data with fixed effects; comparing beta estimates, looking at residuals.
4	Discrete outcomes regression models across countries: binary logistic regression analysis	Lecture – 90 minutes Basics of binary logit models: how it works, MLE estimates, and assumptions, logit estimates, odds ratios and predicted probabilities; avoiding some tricky mistakes in comparing logit across countries. Lab – 90 minutes Running a binary logit model analysis based on crossnational survey data.
5	Discrete outcomes regression models across countries: multinomial logistic regression analysis	Lecture – 90 minutes Basics of multinomial logit models: how it works, MLE estimates, and assumptions, logit estimates, odds ratios and predicted probabilities; the critical issue of fixing the reference categories in multinomial logit analysis. Lab – 90 minutes Running a multinomial logit model analysis based on crossnational survey data; looking at the effects of changing the reference category.
6	Loglinear models across countries: testing for the invariance of odds-ratios across countries	Lecture – 90 minutes Basics loglinear models analysis: how it works, MLE estimates and assumptions, relative chances and odds-ratios estimates across countries; the 'no three-ways' interaction hypothesis and the logic of 'constant' odds-ratios across countries. Lab – 90 minutes Running a loglinear models analysis based on crossnational surveys data; testing different hypotheses about the 'country effect', comparing models.
7	Scaling across countries: how to construct scales from crossnational surveys data?	Lecture – 90 minutes Recall about scales analysis and assumptions: from additive/Likert scales to factor analysis/PCA analysis. Lab – 90 minutes Constructing attitudinal scales from a crossnational survey database.
8	Scaling across countries: how to test for crossnational invariance of scales?	Lecture – 90 minutes What is 'factorial invariance'? What are the most important issues of this hypothesis for crossnational survey analysis? Why is this so important for comparative survey analysis? Lab – 90 minutes Testing for crossnational invariance using Tucker phi coefficient; introduction to structural equations models.
9	Contextualising regression analysis and linking micro and macro levels: multilevel analysis for crossnational analysis	Lecture – 90 minutes Basics of multilevels regression analysis: what it is, how to read a multilevel analysis output. Lab – 90 minutes Running a multilevel (2-level) regression analysis.
10	Conclusions: crossnational analysis clustering of countries – introduction to cluster analysis and discontinuity between countries	Classroom – 3 hours Concluding lecture Student presentations

Day	Readings
1	Alan Agresti, Barbara Finlay Statistical Methods for Social Sciences 4th edition Chapters 5, 6, 7 and 9
2	Paul M. Kellstedt, Guy D. Whitten The Fundamentals of Political Science Research Chapter 8, extracts Cambridge University Press, 2013
3	Paul M. Kellstedt, Guy D. Whitten The Fundamentals of Political Science Research Chapters 9 and 10, extracts Cambridge University Press, 2013 Alfred DeMaris Regression with Social Data: Modeling Continuous and Limited Response Variables pp.148–154 (compulsory) Wiley, NJ, 2004
4	Paul M. Kellstedt, Guy D. Whitten The Fundamentals of Political Science Research Chapter 11, extracts Cambridge University Press, 2013
5	Guy D. Whitten and Harvey D. Palmer Heightening Comparativists' Concern for Model Choice: Voting Behavior in Great Britain and the Netherlands American Journal of Political Science, Vol. 40, No. 1 (Feb 1996) pp.231–260
6	Jacques Hagenaars Loglinear Models with Latent Variables Quantitative Applications in the Social Sciences Series Sage Publications, 1993
7	Pennings, Paul, Hans Keman, and Jan Kliennijenhuis Doing Research in Political Science: An Introduction to Comparative Methods and Statistics Chapter 4 London, Sage Publications, 2006
8	James Georgas, Kostas Mylonas, Aikaterini Gari & Penny Panagiotopoulou Families and Values in Europe In: Wil Arts and Loek Halman (ed). European Values at the Turn of the Millennium Brill, Leiden, 2004 Jia He, Fons J.R. van de Vijver Research methods In: David Matsumoto, Hyisung C. Hwang The Handbook of Culture and Psychology, 2nd edition Oxford, Oxford University Press, 2019
9	Kreft, Ita, and Jan De Leeuw Introducing Multilevel Modeling Chapter 1 London: Sage, 1998 Bickel, Robert Multilevel Analysis for Applied Research: It’s Just Regression! extracts from Chapters 1–3 New York: Guilford Press, 2007
10	No readings

Install the app