Fig. 1

Latent growth model assuming missing data to be MAR (solid black) to calculate differences between intervention and control group, amended by three MNAR sensitivity analyses: Diggle-Kenward selection model (dashed grey), Wu-Carroll shared parameter model (solid grey) and pattern mixture models (dotted grey). Note. The factor loadings from the latent slope factor to the repeated outcome variables correspond to the time between measurement occasions (0.1 = 3 months). The arrows and factor loadings for the latent intercept and quadratic factor were omitted for clarity. ε₀ to ε₄ represent residual variances of the repeated outcome measures. m1 to m4 represent missing data indicators indicating for every participant if the outcome at the respective time point t1 to t4 was observed (m = 1) or not (m = 0)