Confounding

A confounding variable (also confounding factor, a confound, or confounder) is an extraneous variable in a statistical model that correlates (directly or inversely) with both the dependent variable and the independent variable. When the dependent and independent variable have no connection (no causal pathway), the intervention of the confounding variable can produce a spurious relationship between both of them, producing errors in estimation.

Confounding variables is a concern is complex systems, as they are social sciences, psychology or others. Statistics have to deal with that and the experiments have to be adapted to reduce the possibility to have confounding variables not detected in our system by using control testing. In order to control the confounding effects there are some experimentation techniques as:

• Case-control studies: assign confounders to both groups, cases and controls, equally. Both groups have only a variation on the supposed causal variable and they are tested a posteriori.
• Cohort studies: apply in longitudinal studies in which there is a passive follow-up of a group of people and a documentation of relevant characteristics or events related to this group of people.
• Double blinding: neither the experimenter, neither the trial population know to what group they belong.
• Randomized controlled trial: A method where the study population is divided randomly in order to mitigate the chances of self-selection by participants or bias by the study designers. Before the experiment begins, the testers will assign the members of the participant pool to their groups (control, intervention, parallel), using a randomization process such as the use of a random number generator.
• Stratification: the study population sampled is then stratified, ensuring that each group has at least some representativity of the variables under study.
• Controlling for confounding: by measuring the known confounders and including them as covariates is multivariate analyses such as regression analysis. Multivariate analyses reveal much less information about the strength or polarity of the confounding variable than do stratification methods.

See also

Statistics, Hierarchy of evidence

Papers

• Pearl, J. (January 1998). Why there is no statistical test for confounding, why many think there is, and why they are almost right. UCLA Computer Science Department, Technical Report R-256.

Books

• Montgomery, D. C. (2001). Design and Analysis of Experiments (Fifth ed.). Wiley.