centrality
Definition
Centrality refers to a set of graph-theoretic metrics used within network psychometric models to assess the relative importance or position of individual nodes, typically observable symptoms, within a network structure. In psychopathology research, identifying central symptoms has direct implications for designing and evaluating psychological interventions, on the assumption that targeting highly central nodes will produce the greatest therapeutic benefit. These interpretations depend on model accuracy, because omitted variable bias and unmodeled measurement error can distort estimated edge weights and network density in ways that cause researchers to misidentify which symptoms are actually most central. When a network model is misspecified through excluded variables or unreliable items, spurious edges may appear and true connections may be suppressed, both of which alter centrality estimates and can point intervention strategies toward the wrong targets.
Sources: Henry & Ye (2024)
Related Terms
- network psychometrics (1 shared article)
- measurement error (1 shared article)
- omitted variable bias (1 shared article)
Applications
Centrality and Omitted Variable Bias
Omitted variable bias poses a direct threat to valid centrality estimation because variables excluded from a Gaussian Graphical Model can produce spurious edges or suppress true ones, distorting the overall network structure on which centrality calculations depend. When the network structure is misrepresented in this way, the nodes identified as most central may not reflect the actual importance of those symptoms in the underlying disorder, with potentially harmful consequences for intervention design.
Sources: Henry & Ye (2024)
Centrality and Measurement Error
Measurement error that is left unmodeled tends to inflate edge weights and overestimate network density, making the construct appear more interconnected than it actually is. Because centrality metrics are computed from the pattern and magnitude of edges in the network, this inflation directly biases which nodes are ranked as most central, increasing the risk that intervention strategies target symptoms whose apparent importance is an artifact of unreliable measurement.
Sources: Henry & Ye (2024)
