mgcv miscellanea

August 5th, 2018

We've done a lot today. But there's more!

Some other topics we can touch on. Ask as about:

  • More nuanced prediction: variation in smooth shape and derived values
  • A beastiary of smooth types
  • Random effects and heirarchical GAMs (GAMMs!): Three different ways
  • Different slopes for different folks: slope-random effect interactions
  • New approaches to full Bayesian GAMs with Stan or JAGS
  • The hidden gems in mgcv help

Exercises

In the project folder you will find a series of .Rmd files with additional excercises:

  • example-spatial-mexdolphins.Rmd: An extended version of this morning's dolphin excercise. Includes using quantile residuals, another checking tool, which you'll find in the code_snippets/ folder.

  • example-spatio-temporal-data.Rmd: Using tensor smooths to model and separate interactions between smooths that may operate at different scales. Also random effects.

  • example-linear-functional.Rmd: For when your \( y \) outcome variable is dependent on a nonlinear or weighted average of multiple \( x \) variables.

Exercises (2)

  • example-nonlinear-timeseries.Rmd: Time series analysis with decomposition into components. Cyclic (seasonal) smooths, temporal autocorrelation using gamm()

  • example-bivariate-timeseries-and-ti.Rmd: Extending time series analysis to interaction between the components at different time scales

  • exampled-forest-health.Rmd: Modeling ordered categorical outcomes, and discrete spatial units with Markov random fields