What Confounding is
In statistics, confounding is a phenomenon that occurs when the relationship between two variables is obscured due to the presence of a third, extraneous variable. This phenomenon is also known as “spurious correlation” or “omitted variable bias”.
For example, let’s say we’re studying the relationship between the amount of time someone spends studying and their grade in a class. It may appear that studying more leads to better grades, but what if there’s an extraneous variable such as intelligence that’s actually causing the correlation? In this case, studying more and better grades would be confounded by intelligence.
Steps for Confounding:
-
Identify the variables that are being studied.
-
Determine whether there could be any confounding variables that could be obscuring the relationship between the two variables.
-
If there is a potential confounding variable, try to control for it in the study by isolating it in some way.
-
Analyze the data and compare the results before and after controlling for the confounding variable.
-
If the results changed significantly, this could indicate that a confounding variable was present.
Examples
-
A study examining the relationship between smoking and lung cancer may be confounded by the fact that some of the smokers in the sample also consume alcohol, which is another risk factor for lung cancer.
-
A study examining the relationship between diet and cholesterol levels may be confounded by the fact that some people in the sample also exercise regularly, which could be influencing their cholesterol levels.
-
A study examining the relationship between income and mental health may be confounded by the fact that some of the people in the sample also have access to better healthcare, which could be influencing their mental health.