DIW Berlin: Transition and Duration Models

Transition and Duration Models

Termin

30. Oktober - 3. November 2023

Ort

DIW Berlin

Sprecher*innen

Christian Schluter

Zum Kalender hinzufügen

Objectives

Students will study models of transitions and durations, and learn how to estimate these using realworld data.

Outline

This course is an introduction to modelling transitions into a state of interest (such as the transition into employment from unemployment) and durations (such as unemployment, survival of patients after medical treatment or firms after a financial crash). We start with the basic building blocks (Poisson processes, Markovian transitions, hazard models). Since duration data might be censored (individuals might still be in the state of interest at the end of the observation window), classic ordinary least squares (OLS) is invalid, and we develop appropriate methods for estimation. Unobserved heterogeneity introduces fundamental identification challenges.

Throughout this course, all methods will be illustrated using examples in R and Python, and we will consider several papers from the established empirical literature.

Date and Time:
30 October 2023 to 3 November 2023

Mon, Tue, Wed, Fri:
09:30 – 13:00h

Thu:
13:00 – 16:30h

Place:
Mon, Tue: Karl Popper, Room No. 2.3.020
Wed: Ferdinand Friedensburg, Room No.2.3.001
Thu: France D. Blau, Room No. 3.3.002C
Fri: Elinor Ostrom Hall, Room No. 1.2.019

Topics and Ressources

Topics

Literate Programming, Replicability and Reproducibility
Introduction to Counting Processes: The Poisson Process
- Exercise Set 1
- Empirical exercise (Poisson regression): Bike rides
- Empirical exercise (Poisson regression): Doctor visits
- Empirical exercise (PWE): Criminal recidivism
Introduction to Markov Chains
- Exercise Set 2
- Coding: Unemployment transitions
- Coding: Google’s PageRank
Duration and Survival Analysis: Hazard Models
- Exercise Set 3
- Empirical exercise: Criminal recidivism
- Empirical exercise: Time to publication
- Empirical exercise: Divorce in the US
- Empirical exercise: Replication of Pebley and Stupp (1987) “Child Mortality in Guatemala”
- Empirical exercise: The Cancer Genome Atlas (TCGA) data
The Mixed Proportional Hazard (MPH) Model
The Proportional Hazard (PH) Model and Grouped Data
Cox Partial Likelihood
Machine Learning and Survival Analysis: Random Survival Forests.
- Empirical exercise: Training a Cox model
- Empirical exercise: Random Survival Forests

Ressources

Textbook support:
- James et al. (2021): An Introduction to Statistical Learning, Chapter 11.
- Wooldridge (2002), Chapters 19 and 20,
- van den Berg, Chapter 55, Handbook of Econometrics.
Resources on the Web for R:
- (general) web books: R4DS (https://r4ds.had.co.nz/), R Graphics Cookbook (https://rgraphics.org/)
- Survival analysis:
  - the core library: survival (https://cran.r-project.org/web/packages/survival/vignettes/survival.pdf), ggplot for survival curves: survminer (https://cran.r-project.org/web/packages/survminer/vignettes/Informative_Survival_Plots.html), a nice vignette for eha (https://cran.r-project.org/web/packages/eha/vignettes/parametric.html)
  - advanced (ML): Random survival forest: web (https://www.randomforestsrc.org/cheatsheets.pdf),
    vignette (https://www.randomforestsrc.org/articles/getstarted.html).
Resources on the Web for Python:
- lifelines vignettes (https://lifelines.readthedocs.io/en/latest/index.html), scikit-survival
  (https://scikit-survival.readthedocs.io/en/stable/user_guide/understanding_predictions.html)
Resources on the Web for Markdown and Latex:
- cheat sheets: Markdown (https://github.com/adam-p/markdown-here/wiki/Markdown-Cheatsheet)

About the instructor

Christian Schluter is a Professor of Economics at Aix-Marseille School of Economics. His research is in the wider field of empirical labour economics / econometrics, and has been published in e.g. The Review of Economic Studies, The Review of Economics and Statistics, International Economic Review and Journal of Econometrics.

Registration

If you want to join this short course, please register with the Graduate Center on a first-come, first-serve basis: gradcenter@diw.de