Figure 7.
The dynamic path model with direct and indirect effects. This figure shows a dynamic path model with four covariates (genes) X1,X2,X3 and X4 illustrating how the strength of the direct and indirect effect of X1 on survival dN(t) is calculated. Interactions between the covariates are shown as arrows (edges). Each γij(t) is an ordinary least squares regression coefficient at time t, and each βj(t) is an additive hazard regression coefficient at time t. The strength of the direct effect of X1 on dN(t) is given by β1(t). There are two indirect paths from X1 to dN(t): one through X2 and one through X3 and X4. The strength of the indirect effect of X1 on dN(t) is therefore given by γ12(t)β2(t)+γ13(t)γ34(t)β4(t). The strength of the total effect is the sum of the strength of the direct and indirect effect, that is, β1(t)+γ12(t)β2(t)+γ13(t)γ34(t)β4(t).