Figure 5.

Longitudinal mediator–outcome process models with lagged b paths.

Numbers are estimate (standard error). R: randomised treatment group; M₀, M₁₂, M₂₄, M₅₂: mediator measurements taken at baseline, 12 weeks, 24 weeks and 52 weeks post-randomisation; Y₀, Y₁₂, Y₂₄, Y₅₂: outcome measurements taken at the same time points; FM₀, FM₁₂, FM₂₄, FM₅₂: true mediator scores at baseline, 12 weeks, 24 weeks, 52 weeks post-randomisation; FY₀, FY₁₂, FY₂₄, FY₅₂: true outcome scores at the same time points. Paths between observed and latent variables for both the mediator and the outcome process all have factor loadings = 1.