Figure 2

HR(t) configuration chosen. Increasing then decreasing H R(t) configuration, for each Z profile and on average, without censoring and with ${\mathit{\beta }}_{12}^{{}^{\prime }}=\left(-0.2,-0.4,-0.8\right)$, ${\mathit{\beta }}_{13}=-{\mathit{\beta }}_{12}$ and ${\mathit{\beta }}_{23}=-{\mathit{\beta }}_{13}$. This figure displays the theoretical estimations of $\overline{\mathit{\text{HR}}}\left(t\right)$ called “mean”, ${\overline{\mathit{\text{HR}}}}_a\left(t\right)$ called “adjusted mean” and H R _i (t) in the eight prognostic profiles. The profile ${\mathbf{Z}}^{{}^{\prime }}=(0,0,0)$ at time t=0 is the profile with the better prognosis; the profile ${\mathbf{Z}}^{{}^{\prime }}=(1,1,1)$ has the worse prognosis, and the 6 others an intermediate prognosis. In this particular configuration chosen, where ${\mathit{\beta }}_{23}=-{\mathit{\beta }}_{13}$, $\overline{\mathit{\text{HR}}}\left(t\right)\simeq {\overline{\mathit{\text{HR}}}}_a\left(t\right)$ and their values are so close that the difference between them is not visible in this figure.