GLM example for proportion data

1 Lecture

Slides in full screen

2 Exercise

Conduct a detailed analysis of the study on survival of prairie dogs in winter in the dataset 08_prairiedogs.csv.

  1. Graphical data exploration
    1. Create appropriate figures to explore all potential two-way interactions.
    2. Create appropriate figures to explore all potential main effects.
  2. Model and test all possible two-way interactions
    1. Fit a model with all main effects and all two-way interactions
    2. Test the two-way interactions using Likelihood-Ratio-Tests
  3. Model and test the main effects
    1. Fit a model with all main effects
    2. Test the main effects using Likelihood-Ratio-Tests

2.1 Purpose of the analysis

  • Discuss if the purpose of this study is exploration, prediction, or inference!
  • What does this imply for the interpretation of your findings?
  • How would you report your findings in a report?

2.2 Extra

  • Create a figure with the data and the model predictions of the best model that you found.

  • Check if you get the same results when you define your response variable as

    1. Proportion plus weights
    2. Two-column table (see slide 13)

For the visualisation of interactions between two categorical variables boxplots with fill colour are a good choice. Use aes(..., fill = your_variable)

Solutions