????????????????????????????????????

Dr Anja Neundorf is an Associate Professor in Politics and Research Methods at the University of Nottingham. She previously held a Post-doctoral Prize Research Fellowship at Nuffield College, Oxford and received her PhD from the University of Essex. Her research interests lie at the intersection of political behaviour, research methods, and comparative politics. Her research has been published in the Journal of Politics, Public Opinion Quarterly, and Social Forces.

Course Content
This is an introduction to longitudinal data analysis on an applied level using Stata. The focus of the course is on data management and analysing micro panel data.

The module will begin by discussing the advantages (and limitations) of panel data, and will show how to handle and describe a panel dataset. We will then cover linear regression techniques: fixed and random effects models as well as hybrid models, and simple dynamic regression models as well as growth curve modelling. Next we will look at non-linear models, such as the random effects probit and fixed effects logit, which are used to deal with discrete variables. The last part of the course will then turn to the analysis of repeated cross-sectional data, focusing on age, period, cohort (APC) analysis, a technique to identify over time societal dynamics. We will focus on the conventional methods as well as the estimation of hierarchical APC models.

Following each lecture, participants will work through practical examples in the computer lab using the Stata statistical package and diverse datasets, e.g. British Household Panel Survey, the German Socio Economic Panel and the British Social Attitude Survey. The focus of the module is applied, but some maths will be used to formalise theoretical concepts. Note that the module does not cover specific techniques for macro panels (e.g. data on countries over time) or panels with small numbers of cross-sectional units but many time points. The module does not cover survival (event history or duration) or time-series analysis. The module also does not cover household dynamics, but focuses on individuals within households.

Course Objectives

To develop the skills necessary to understand and assess the applications of longitudinal data analysis reported in the social science literature; and to enable participants to apply longitudinal data techniques to their own research questions. Further, participants will learn how to create working files of longitudinal data such as household panel studies (e.g. the British Household Panel Survey) or repeated-cross sectional data (e.g. British Social Survey).

Course Prerequisites

Essential requirements for the module are:
• (a) Final year undergraduate level knowledge of linear regression methods (OLS regression) and some familiarity with issues like sample selection and endogeneity. Some experience of multilevel modelling and non-linear methods like probit and logit would also be useful. Remedial reading for these topics is Verbeek, chapters 1–3, 5 and 7 (see below).
• (b) Intermediate level proficiency in Stata: familiarity with basic commands and experience of writing Stata do files.

Essential reading

Andreß, Hans-Jürgen, Golsch, Katrin, and Schmidt, Alexander W. 2013. Applied Panel Data Analysis for Economic and Social Surveys. Springer.
• Andreß et al is an intuitive and fairly non-technical guide. It contains a very good treatment of descriptive techniques and key modelling concepts, but does not cover more advanced dynamic models, instrumental variables, or more sophisticated non-linear models.

Rabe-Hesketh, Sophia and Skrondal, Anders. 2012. Multilevel and longitudinal modelling using Stata. Vol. 1: Continues responses. Third Edition. College Station, Texas: Stata Press.
• This is a practical guide to longitudinal analysis, giving intuitive, non-technical instructions to most issues covered in this course. It also gives detailed instructions on how to implement the analyses in Stata.

More technical and specific reading

There is no single text, which covers all the module topics in a way that is accessible to applied researchers (dedicated panel data texts tend to be quite technical). The following recommended books contain useful material. Short descriptions of each book are provided. It is recommended to buy at least one of these books.

Callegaro, Mario, Baker, Reg, Bethlehem, Jelke, Goeritz, Anja S., Krosnik, Jon A., and Lavrakas, Paul J. 2014. Online Panel Research: A Data Quality Perspective. Chichester, West Sussex: Wiley.
• This is an excellent book for everyone planning to work with online panel data. It focuses on the data quality issues though and less on analytical tools.

Diggle, Peter J., Heagerty, Patrick, Liang, Kung-Lee and Zeger, Scott L. 2002. Analysis of Longitudinal Data. Oxford: Oxford University Press.
• Technical book for the more advanced readers.

Frees, Edward W. 2004. Longitudinal and Panel Data: Analysis and Applications in the Social Sciences. New York, NY: Cambridge University Press. (freely available here: http://instruction.bus.wisc.edu/jfrees/jfreesbooks/Longitudinal%20and%20Panel%20Data/Book/Chapters/FreesFinal.pdf)
• Medium-level book with a good combination of mathematical and statistical fundamentals and substantive applications from across the social sciences. Comprehensive coverage of topics.

Hsiao, C. 2003. Analysis of Panel Data (Econometric Society Monographs). (2nd ed.) Cambridge University Press.
• Classic book in panel data analysis. Comprehensive coverage of topics, including more advanced topics.

Rabe-Hesketh, Sophia and Skrondal, Anders. 2012. Multilevel and longitudinal modelling using Stata. Vol. 2: Categorical Responses, Counts, and Survival. Third Edition. College Station, Texas: Stata Press.
• This is a practical guide to longitudinal analysis, giving intuitive, non-technical instructions to most issues covered in this course. It also gives detailed instructions on how to implement the analyses in Stata. This is not essential for those not working with categorical data.

Verbeek, Marno. 2012. A Guide to Modern Econometrics. (4th ed.). Wiley.
• Participants should be comfortable with the material in ch. 1–3, 5 and 7 before the course. Ch. 10 covers panel data. Verbeek, chapter 10, is more formal, but still accessible, introduction to both basic and more advanced models.

Yang, Yang, Land, Ken C., 2013. Age-Period-Cohort Analysis: New Models, Methods, and Empirical Applications. CRC Press: Taylor and Francis Group.
• Comprehensive book on APC analysis.

Course Content and Teaching Method

This is an introduction to longitudinal data analysis on an applied level using Stata. The focus of the course is on data management and analysing micro panel data.

The module will begin by discussing the advantages (and limitations) of panel data, and will show how to handle and describe a panel dataset. We will then cover linear regression techniques: fixed and random effects models as well as hybrid models, and simple dynamic regression models as well as growth curve modelling. Next we will look at non-linear models, such as the random effects probit and fixed effects logit, which are used to deal with discrete variables. The last part of the course will then turn to the analysis of repeated cross-sectional data, focusing on age, period, cohort (APC) analysis, a technique to identify over time societal dynamics. We will focus on the conventional methods as well as the estimation of hierarchical APC models.

Following each lecture, participants will work through practical examples in the computer lab using the Stata statistical package and diverse datasets, e.g. British Household Panel Survey, the German Socio Economic Panel and the British Social Attitude Survey. The focus of the module is applied, but some maths will be used to formalise theoretical concepts. Note that the module does not cover specific techniques for macro panels (e.g. data on countries over time) or panels with small numbers of cross-sectional units but many time points. The module does not cover survival (event history or duration) or time-series analysis. The module also does not cover household dynamics, but focuses on individuals within households.

Day-by-Day Program

Day 1: Basics – Longitudinal Data
· What are longitudinal data (examples)?
· Why use panel data? Advantages and challenges.
· Handling panel data in Stata – some basic commands.
· Patterns of observations in panel data (non-response and attrition)
· Data management: Building a working file using the British Household Panel Survey

Core Readings: Andress et al. 2013. Chapter 1-2.2

Day 2: Describing panel data
· Within and between variation
· Transitions
· Sequencing
· Survival or event history analysis
· An example: Studying poverty using panel data
· Using weights

Core Readings: Andress et al. 2013. Chapters 2.3-2.5; 3.3

Day 3: Modelling panel data – Introduction
· Different types of independent variables
· Modelling panel data – overview
· Challenges of panel data modelling
– Statistical dependencies: Serial-correlation in panel data
– Unobservables
– Measurement error bias

Core Readings: Andress et al. 2013. Chapter 3.

Day 4: Modelling continuous-level dependent variables – Levels (1)
· Pooled Ordinary Least Squares Robust standard errors
· Fixed effects regression
– Least Squares Dummy Variables (LSDV)
– Within regression (time demeaning)
· Between-group regression

Core Readings: Andress et al. 2013. Chapters 4.1.-4.1.2.1

Day 5: Modelling continuous-level dependent variables – Levels (2)
· Review within-group (FE) models
· Random effects regression
· Comparing, pooled OLS, FE, BE and RE regression
· Hybrid models
· Test statistics
· Unbalanced data

Core Readings:
· Andress et al. 2013. Chapter 4.1.2.2-4.1.3.
· Bell, A. & Jones, K. 2015. Explaining Fixed Effects: Random Effects Modeling of
Time-Series Cross- Sectional and Panel Data. Political Science Research and Methods, 3(1): 133-153. [included in reading pack]

Day 6: Modelling categorical dependent variables – Levels
· Binomial Logistic and Probit Regression – An overview – Why do we need to use this for categorical DVs? – How to interpret the results of logistic regression? – Maximum-Likelihood Estimation
· Modelling Binary DVs in Panel Data
– Within-group (FE) estimation
– Random effects estimation
– Choosing between FE and RE estimation

Core Readings: Andress et al. 2013. Chapter 5.1.

Day 7: Modelling continuous-level dependent variables – Change
· Analysis of change using impact functions
· Dynamic models using lagged DV
· Growth curve models

Core Readings:
· Andress et al. 2013. Chapter 4.2.1-4.2.2.
· Hox, J. & Stoel, R. 2005) Multilevel and SEM Approaches to Growth Curve Mod-eling. Encyclopaedia of Statistics in Behavioral Science, 3: 1296-1305. [included in reading pack]

Day 8: Modelling categorical dependent variables – Change
· Review: Modelling categorical-level dependent variables – Level
· Unobserved heterogeneity versus state dependence
· Markov Chain modelling:
– General overview
– Example: Mover-Stayer Model

Core Readings:
· Neundorf, A., Stegmueller, D. & Scotto, T. 2011. The individual level dynamics of bounded partisanship. Public Opinion Quarterly, 75 (3): 458-482. [included in reading pack]

Day 9: Working with Repeated Cross-Sectional Data – Intro APC analysis
· What is age, period, cohort analysis about?
– Understanding APC effects: An Example using turnout – The APC “Conundrum”
– Data requirements
· Exploring APC effects
– The age-by-time period table – Using graphs

Core Readings:
· Neundorf, A. & R. Niemi. 2014. Beyond political socialization: New approaches in age, period, cohort analysis. Electoral Studies, 33: 1-6. [included in reading pack]
· Andress et al. 2013: Chapt. 4.2.3.

Day 10: Breaking the APC “conundrum”: Conventional and new methods to conduct APC analysis
· Conventional approaches
– Coefficient constraint approach – Reduced two-factor model – Proxy variables
· New methods and approaches
– Example: Hierarchical Age-Period-Cohort (HACP) models Practical application: Political context & turnout

Core Readings:
· Yang, Y., Land, K.C. 2006. A mixed models approach to age-period-cohort analysis of repeated cross-section surveys: trends in verbal test scores. Sociological Method¬ology 36, 75-97. [included in reading pack]
· Smets, K. and A. Neundorf. 2014. The hierarchies of age-period-cohort research: Political context and generational turnout patterns. Electoral Studies, 33: 41-51. [in¬cluded in reading pack]