Smoothing Demographic Data

In 2017, as part of the International Advanced Studies in Demography at the Max Planck Institute for Demographic Research, I have taught a 20-hours course titled "Smoothing Demographic Data: Flexible Models in Population Studies". Course description from the MPIDR web-page can be found here.

For this course, within a week, I prepared 5 lectures during the morning with associated lab-sessions in R in the afternoon. Specifically I covered the following topics:

• Lecture 1: Introduction to Generalized Linear Model

  • Introduction to Generalized Linear Model

  • Basic linear model

  • Let’s forget about normality

  • The Iterative Weighted Least-Squares algorithm

  • Estimating Poisson model by hand

• Lecture 2: Discrete smoothing

  • The pursuit of smoothness

  • Penalty! (Double) Goal!

  • Optimizing the amount of smoothness

  • Smoothing histograms

  • Including exposures

• Lecture 3: P-splines

  • Let’s forget about linearity

  • B-splines: What’s that?

  • Penalizing the coefficients

  • Smoothing counts

  • Simple extrapolation

• Lecture 4: Extending P-splines

  • Dealing with more covariates

  • Smoothing spatial data

  • Tensor P-splines

  • Shape constraints

  • Calculus with smooth data

• Lecture 5: Introduction to Generalized Additive Models (for Location, Scale and Shape)

  • Presenting mgcv by examples

  • Presenting gamlss by examples