In clustered data, observations share common features not accounted by covariates in the model. This describes?

• A. Clustering
• B. Random sampling
• C. Supervised learning
• D. Time-to-event data

Answer: (A)

Clustering is when observations within the same group share features that aren’t captured by the covariates in the model, making those observations more similar to each other than to observations in other groups. This produces intra-cluster correlation, so the usual assumption of independence of observations is violated. Because of that, standard errors can be biased if clustering isn’t accounted for, leading to overstated precision or incorrect inferences. This description fits clustering, since it emphasizes the shared, unmeasured characteristics within groups (like patients within the same hospital or students within the same classroom).

Random sampling refers to how data are drawn from a population, not the within-group dependence. Supervised learning is about predicting outcomes using labeled data, not the data’s correlation structure. Time-to-event data describes the type of outcome and its analysis, not the presence of grouped similarity.