Structural equation models with standardized path coefficients (β) showing the MRI-depression and MRI-cognition path in A. the direct effect model, B. the mediation model, and C. the final model. In ‘A’ model, white matter lesion (WML) predicted both depressive score (β = 0.30, SE = 0.08, p < 0.001) and cognition (β = −0.25, SE = 0.09, p = 0.01). In ‘B’ model, the latent variable ‘frontal cortex’ was added as a mediator. Note that frontal cortex (β = − 0.02, SE = 0.10, p = 0.85) did not predict depressive score and WML volume (β = −0.003, SE = 0.09, p = 0.97) did not predict cognition directly. Therefore, in ‘C’ Model, we removed frontal cortex in MRI-depression path and the direct path from WML to cognition in MRI-cognition path. Note that WML predicted depressive score directly but predicted cognition through frontal cortex. Circles represent latent variables, and rectangles represent measured variables. Estimated error terms were omitted to simplify the figures. *p < 0.05.
Lee et al. BMC Neurology 2012 12:83 doi:10.1186/1471-2377-12-83