Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Figure 1.

Coefficient estimates for pre-selected markers obtained from 100 simulation runs.

The marker combinations were optimized via gradient boosting based on training samples of size (left) and (right). Boxplots represent the empirical distribution of the resulting coefficients. Only markers to had an actual effect on the survival time.

More »

Figure 1 Expand

Figure 2.

Simulation results for the discriminatory power obtained via the proposed -index boosting approach and competing Cox-based estimation schemes.

The marker combinations were optimized via the different approaches based on training samples of size (left) and (right). Boxplots represent the empirical distribution of the resulting on corresponding test samples. The dotted line corresponds to the discriminatory power resulting from the true combination of predictors with known coefficients.

More »

Figure 2 Expand

Table 1.

Results of the simulation study.

More »

Table 1 Expand

Table 2.

Simulation results for different values of the smoothing parameter.

More »

Table 2 Expand

Figure 3.

Comparing the discriminatory power of biomarker combinations to predict the time to distant metastases resulting from the proposed -index boosting approach with competing estimation schemes.

The plot on the left refers to the Desmedt et al. data, whereas the plot on the right presents results from the van de Vijver et al. data. All biomarker combinations were optimized via the corresponding algorithms based on the same 100 learning samples. Boxplots represent the empirical distribution of the resulting on corresponding test samples. The dotted line corresponds to the median -index resulting from the new approach.

More »

Figure 3 Expand