Modelling continuous outcomes

Linear regression modelling is one of the most ubiquitous methods applied in the social sciences. It applies to cases where the phenomenon we aim to understand and describe (model) is ideally measured on a continuous numeric scale. This will be our “dependent” or “outcome” variable (our “explanandum”, the variable we want to explain). Regression modelling allows us to quantify the association between one or several explanatory (“independent”) variables and our outcome variable. This allows us to summarise associations and draw comparisons in our data more efficiently and to establish testable hypotheses about potential causal relationships. In the workshop we will scratch the surface of this versatile and foundational statistical method.

Essential readings

Access links through Canvas - Newcastle University login required

1. Spiegelhalter (2020) The Art of Statistics:
   • CHAPTER 5 | Modelling Relationships Using Regression
2. Goss-Sampson (2025) Statistical Analysis in JASP:
   • REGRESSION (pp. 75-88)

Application:

• Wilkinson & Pickett (2010) The Spirit Level, Chapter 4
• Delhey & Newton (2005) Predicting Cross-National Levels of Social Trust
• Österman (2021) Can We Trust Education for Fostering Trust?
  • Plus, the Electronic supplementary material associated with the article.

Further readings

Statistics:

• Kranzler (2022) Statistics for the Terrified:
  • Chapter 11 (“Correlation Coefficients”)
  • Chapter 12 (“Linear regression”)
• Çetinkaya-Rundel and Hardin (2024) Introduction to Modern Statistics:
  • Chapter 7 (“Linear regression with a single predictor”)
  • Chapter 8 (“Linear regression with multiple predictors”)

Advanced topics:

• Clark, William Roberts, and Matt Golder (2023) Interaction Models: Specification and Interpretation. Cambridge: Cambridge University Press.