A confounder is a factor — either observed or unobserved — which affects the relationship between two variables being studied. Not accounting for the potential effect of a confounder may lead to incorrect conclusions about the relationship between the variables.


A confounder — also called a confounding variable/factor — is a third variable that, because of its relationship to the two variables being studied, can affect the relationship between those variables. Let’s consider an example to better illustrate the concept.

Imagine that a researcher is examining the relationship between exercise and body weight by analyzing data from a group of exercising participants and from a group of nonexercising participants. Initially, the researcher finds that the exercising participants tend to have a lower body weight than the nonexercising participants, leading the researcher to conclude that exercise is related to a lower body weight. However, upon further analysis, the researcher discovers that the exercising participants also tend to have healthier eating habits and consume fewer calories than the nonexercising participants. In this case, diet acts as a confounder by potentially magnifying the strength of the relationship between exercise and lower body weight.

To address this issue, the researcher can collect dietary data from the participants and make statistical adjustments to account for the potential effect of diet. By doing so, the researcher can better assess the true relationship between exercise and body weight. However, there is still the possibility of residual and/or unmeasured confounding.

What is residual confounding?

Residual confounding refers to the influence of a confounding variable on the relationship between two other variables even after efforts to control for it. It generally arises because of limitations in the methods used to control for confounding. In the example given above, the researcher collected dietary data from the participants to account for the potential effect of their diet. Residual confounding could occur if the dietary data collected did not accurately capture all relevant aspects of the diet that could confound the relationship between exercise and body weight. For example, dietary information may have been collected using a 24-hour dietary recall, which has limited ability to describe a person’s typical diet.

What is unmeasured confounding?

Unmeasured confounding refers to the presence of confounding variables that were not considered by the researchers, and were therefore, not measured or accounted for. In the example given above, a possible variable that could introduce unmeasured confounding is socioeconomic status. Because socioeconomic status may be related to both exercise habits and body weight, failure to account for its potential confounding effects can result in unmeasured confounding.