Instrumental Variables: Uses and Limitations

Observational studies are limited by the issue of confounding. For example, light-to-moderate drinkers tend to have lower risk of cardiovascular disease than nondrinkers. This association could be due to a causal effect of alcohol on the body, or it could be due to other  attributes of light-to-moderate drinkers, such as high socioeconomic status or the ability to practice moderation across a range of behaviors. Statistical adjustment only imperfectly addresses confounding [1]: Confounders such as socioeconomic status may be crudely measured, resulting in residual confounding; plus, some confounders—such as intangible attributes like the ability to practice moderation—may remain unmeasured.
Source: PM and R - Category: Rehabilitation Authors: Tags: Statistically Speaking Source Type: research