UE291 - Inferential and causal methods for the Social Sciences [SOCSCI 102]

Intervenant·e·s

• Rita Abdel Sater [référent·e]   doctorante, ENS

This course will provide students with a basic knowledge of data analysis to answer questions of cultural, social, economic, and policy interest: estimation, regression and econometrics, prediction, experimental design, randomized control trials, and data visualization. The sessions will be illustrated with examples and papers with a particular focus on topics and interventions drawing on behavioral science and psychology. The course is open to students from all backgrounds and aims at providing them with the tool-box necessary to understand the importance of causal inference, identify the correct methods to draw causal inferences and develop critical thinking.

Ce cours fournira aux étudiants une connaissance de base de l'analyse des données pour répondre à des questions d'intérêt culturel, social, économique et politique : estimation, régression, essais contrôlés randomisés et visualisation des données. Les sessions seront toujours illustrées par des exemples et des articles mettant l'accent sur des sujets et des interventions s'appuyant sur les sciences du comportement et la psychologie. Le cours est ouvert aux étudiants de tous horizons et vise à leur fournir la boîte à outils nécessaire pour comprendre l'importance de l'inférence causale, identifier les bonnes méthodes en analyse de données et développer la pensée critique.

Prerequisite

A basic knowledge of statistics is required. Students with no prior experience with statistics must take the Coursera course “Basic Statistics”.

Introduction

Lecture 1 : Comparing two groups - Null hypothesis testing, Confidence intervals and two-sided tests, Power

Lecture 2 : Categorical association and simple regression -  Categorical association and independence, Chi-squared test, linear regression, Ordinary least squares, goodness-of-fit

Lecture 3 : Multiple regression - Hypothesis testing in linear regressions, multiple regression analysis, omitted variable bias

Lecture 4 : Multiple regression 2 - Interpreting coefficients, nonlinear relationships ( log and quadratic transformation), categorical predictors

Lecture 5 : Multiple regression 3  - Interactions, joint hypothesis testing, categorical outcome variable

Lecture 6 : Assumptions and pitfalls of regression analysis - OLS assumptions, outliers, ANOVA, non-parametric tests

Session 1 : Practice (with Valentin Thouzeau ) 

Session 2 : Practice (with Valentin Thouzeau ) 

Lecture 7 : Introduction to impact evaluation and Randomized controlled trials - Why do we need impact evaluation? When is it used? What is randomization, why and how does it work? How do we determine the sample size? How do we calculate the effect size? What are the different experimental designs used in the literature? What are potential problems and what are the usual solutions? What are some ethical considerations? 

Lecture 8 : Difference-in-Difference approach - When can we use the difference- in-difference approach? What are the theoretical assumptions behind it and how is it applied in practice? 

Lecture 9 : Regression discontinuity design - What is a regression discontinuity? How and when can we use? What is the theory behind this design and what are its limits? What is the difference between a sharp and a fuzzy design? 

Lecture 10 : Matching methods and instrumental variables - What are the different tools for matching methods? What is propensity score matching? When can we use it and what kind of data do we need?  What are the limits of such methods? What is an instrumental variable? What assumptions should such variables satisfy? What are the most used instrumental variables and what are the critiques about them?