Which term describes a covariance structure used in multilevel modelling that is completely general?

Which term describes a covariance structure used in multilevel modelling that is completely general?

• A. Uninformative prior distribution
• B. Unstructured
• C. Upper quartile
• D. Unique variance

Answer: (B)

In multilevel modelling, how the random effects covary is captured by the covariance structure. An unstructured covariance structure allows every variance and every covariance to be estimated freely—there are no imposed patterns or constraints. This makes it completely general, since it can fit any arrangement of relationships among the random effects. Of course, this flexibility comes with a cost: many parameters to estimate, so you need enough data to support it. The other options aren’t covariance-structure terms: an uninformative prior distribution is a Bayesian concept, not a way of specifying how random effects relate; an upper quartile is a percentile; and unique variance isn’t a standard term for describing a covariance pattern.

In multilevel modelling, how the random effects covary is captured by the covariance structure. An unstructured covariance structure allows every variance and every covariance to be estimated freely—there are no imposed patterns or constraints. This makes it completely general, since it can fit any arrangement of relationships among the random effects. Of course, this flexibility comes with a cost: many parameters to estimate, so you need enough data to support it. The other options aren’t covariance-structure terms: an uninformative prior distribution is a Bayesian concept, not a way of specifying how random effects relate; an upper quartile is a percentile; and unique variance isn’t a standard term for describing a covariance pattern.