Figure 2.
The MN-MS model estimates well non-linear effects of covariates x1, x2, and x3 (top panels) and the absence of effects associated with covariates x4, x5, and x6 (bottom panels). True mean response functions are depicted with red line while the estimated relationship are shown with black lines (continuous and dashed lines are the median and point-wise 95% credible intervals, respectively). Circles show the knot locations for each covariate, a priori set to 0.2, 0.4, 0.6, and 0.8 quantile of the corresponding covariate. The displayed response curves are based on one of the 10 simulated datasets and were created by only varying the focal covariate while the other covariates were set to their mean values.