Occurs when the likelihood of an observation is not independent of its relationship with other observations. For example, autocorrelation occurs when a good day for a patient with chronic disease is more likely to follow a "good day" than a "bad day."