Intermediate level · Module 4: Relationships and Regression for Analysis · Lesson 3 of 3
Confounding and Limits of Observational Data
Recognise confounding and selection bias and describe what observational analyses can and cannot conclude.
In this lesson you will
- Identify confounders and selection bias
- Word conclusions appropriately for observational data
Key concepts
confounderselection biasomitted variablecareful wording
The lesson
A confounder affects both the factor and the outcome. Schools chosen for feeding programmes may already be more disadvantaged; comparing them with other schools can make feeding look harmful even when it helps.
Adding measured confounders to a regression helps, but unmeasured ones remain. Observational data rarely proves cause.
Word conclusions carefully: 'associated with' rather than 'causes', and describe which factors were and were not accounted for. Causal designs are covered at the Advanced level.
Step by step
- Draw the possible confounders of the feeding–attendance relationship.
- Add measured confounders to the regression.
- List unmeasured ones.
- Rewrite conclusions with careful wording.
Practice activity
Write the discussion section of the attendance analysis, covering confounding and limits.
Expected result: A discussion that names confounders and words conclusions correctly.
Check your understanding
-
What is a confounder?
Show answer
A factor that affects both the explanatory variable and the outcome.
-
Which wording suits observational findings?
Show answer
'Is associated with', not 'causes'.
Taught live by an instructor — online or in person. Free course lessons are for self-study and do not lead to a certificate — see registration options.