Fig. 7.

Effects from the Puberty Models: Example replicant for underlying (A) and observed (B) trajectories of pubertal development. Distributions of fixed effects are plotted for models where the puberty predictor is sampled from the continuous puberty curves (C) and models where puberty is assessed by the transformed Tanner Stage values (D). This tranformation introduced substantial bias into the age predictor estimates, whereas no bias was introduced when using the underlying values. Only data from the cohort model are shown.