Categories → Causal Inference , Statistics , Study Design , Causal Effect
An intervention is a deliberate action taken to modify or control a factor of interest in order to observe its effect on an outcome.
In the context of causal inference, an intervention is a deliberate action taken to modify or control a variable of interest in order to observe its effect on an outcome. Typically, the goal of an intervention is to establish a causal relationship between the factor being manipulated (the "treatment") and the outcome of interest.
Interventions can be experimental or observational, depending on the design of the experiment or study being conducted.
In an experimental intervention, the factor of interest is manipulated by assigning participants to different treatment conditions randomly (such as a control group and an experimental group), and the outcome is measured for each group. This is the most commonly used interpretation of "intervention".
In an observational intervention, the treatment is not manipulated by the researcher, but rather occurs naturally or as a result of a policy or process change, and is recorded in historical data. The researcher can then observe the effect on the outcome of interest. This is a huge cost and effort saving, but unless correct causal inference methods are applied, the estimated effect is likely to be biased.
Experimental interventions are an important tool for establishing causality because they allow researchers to directly manipulate a factor and observe its effect on an outcome, while controlling for other factors that could confound the relationship. Yet by carefully defining observational interventions, and using causal inference techniques. researchers can draw also valid conclusions about the causal effects of treatments on outcomes, even for observational data.