confounding

A confounding variable (or confounder) in statistics is an extraneous variable that is not accounted for in a study but influences both the explanatory (independent) variable and the response (dependent) variable. This creates a situation where the observed association between the two variables may be misleading, as the confounding variable could be the true cause of changes in the response variable.

Key Points from the Knowledge Base:

1. Definition (Study Design Lab):

   • A confounding variable "affects the results in an unwanted way, making it difficult to determine if the explanatory variables caused changes in the response."
   • Example: In the Strength Shoes® study, athleticism or genetic factors could confound results if participants self-selected their shoe type. More athletic individuals might choose Strength Shoes®, and their inherent ability—not the shoes—could explain better jumping performance.

2. Impact on Studies:

   • In observational studies, confounding variables limit conclusions to associations rather than causation (since there is no random assignment to control for confounders).
   • In experiments, random assignment helps balance confounding variables between groups, but unmeasured confounders (e.g., BMI or genetic factors in the Strength Shoes® lab) can still bias results if not addressed.

3. Example from the Lab:

   • The Strength Shoes® group had a higher mean BMI (25.58 vs. 21.42) and a higher proportion of the jumping-enhancing genetic factor (75% vs. 25%). These confounders could explain differences in jumping ability instead of the shoes themselves.

Why It Matters:

Confounding variables threaten the validity of causal claims. To establish causation, studies must either:

• Control for confounders (e.g., balancing groups by sex/height in the lab), or
• Use random assignment to distribute confounders equally across groups (as emphasized in the applet activity).

Source: Study Design Lab notes and "Confounding Variables" section in the provided knowledge base.