In fields such as epidemiology, social sciences, psychology and statistics, an observational study draws inferences from a sample to a population where the independent variable is not under the control of the researcher because of ethical concerns or logistical constraints. One common observational study is about the possible effect of a treatment on subjects, where the assignment of subjects into a treated group versus a control group is outside the control of the investigator. This is in contrast with experiments, such as randomized controlled trials, where each subject is randomly assigned to a treated group or a control group. Observational studies, for lacking an assignment mechanism, naturally present difficulties for inferential analysis.

Motivation

The independent variable may be beyond the control of the investigator for a variety of reasons:

Types

Degree of usefulness and reliability

"Although observational studies cannot be used to make definitive statements of fact about the "safety, efficacy, or effectiveness" of a practice, they can:

Bias and compensating methods

In all of those cases, if a randomized experiment cannot be carried out, the alternative line of investigation suffers from the problem that the decision of which subjects receive the treatment is not entirely random and thus is a potential source of bias. A major challenge in conducting observational studies is to draw inferences that are acceptably free from influences by overt biases, as well as to assess the influence of potential hidden biases. The following are a non-exhaustive set of problems especially common in observational studies.

Matching techniques bias

In lieu of experimental control, multivariate statistical techniques allow the approximation of experimental control with statistical control by using matching methods. Matching methods account for the influences of observed factors that might influence a cause-and-effect relationship. In healthcare and the social sciences, investigators may use matching to compare units that nonrandomly received the treatment and control. One common approach is to use propensity score matching in order to reduce confounding, although this has recently come under criticism for exacerbating the very problems it seeks to solve.

Multiple comparison bias

Multiple comparison bias can occur when several hypotheses are tested at the same time. As the number of recorded factors increases, the likelihood increases that at least one of the recorded factors will be highly correlated with the data output simply by chance.

Omitted variable bias

An observer of an uncontrolled experiment (or process) records potential factors and the data output: the goal is to determine the effects of the factors. Sometimes the recorded factors may not be directly causing the differences in the output. There may be more important factors which were not recorded but are, in fact, causal. Also, recorded or unrecorded factors may be correlated which may yield incorrect conclusions.

Selection bias

Another difficulty with observational studies is that researchers may themselves be biased in their observational skills. This would allow for researchers to (either consciously or unconsciously) seek out the information they're looking for while conducting their research. For example, researchers may exaggerate the effect of one variable, or downplay the effect of another: researchers may even select in subjects that fit their conclusions. This selection bias can happen at any stage of the research process. This introduces bias into the data where certain variables are systematically incorrectly measured.

Quality

A 2014 (updated in 2024) Cochrane review concluded that observational studies produce results similar to those conducted as randomized controlled trials. The review reported little evidence for significant effect differences between observational studies and randomized controlled trials, regardless of design. Differences need to be evaluated by looking at population, comparator, heterogeneity, and outcomes.