Three multivariate models were applied to evaluate the contribution from standard clinical parameters alone (Model 1), from standard parameters plus genomic proliferation (Model 2), and from standard parameters plus proliferation and the intrinsic subtype (Model 3).
The log-rank chi-square test statistic for the separation of the survival curves of the high and low risks groups was significantly larger using the model including pCR than the model including preoperative available parameters alone (p < 0.0001, chi test).
A comparison of the log-rank statistics for the differences between the cross-validated Kaplan Meier curves for the high and low risk groups revealed a better separation by the model including histopathologic parameters than by the model including preoperative available parameters alone (p = 0.017, chi test).
Moreover, the predictive power of CCNB2 in addition to the clinicopathological parameters model was slightly higher compared to the lower C-index of the model including all clinicopathological parameters alone.
In fact, as demonstrated recently, AIC and BIC are blind to how much the estimated parameters are dependent on each other, both a priori and a posteriori (Stephan et al. 2009), whereas the F criterion can take into account such interdependency and thus does not penalize models on the number of parameters alone.
Perturbations in the model parameters θD, θN, pD and pN, alone and in combinations, clearly affected the solutions in systematic ways, while variations in μ (the ratio between the degradation rates of Delta and Notch) and in the initial conditions had fairly small effects.
We discover that the model predictions can be accurately characterized by two independent parameters alone: The initial fraction of methane and the molar mass of the remaining non-methane part.
The true accuracy of our multivariable gene expression models most likely lies between the median values for the LASSO selected models and the five/six gene signature models, which is comparable to the accuracy based on clinical parameters alone.
Survival and detection probability estimates of smolt populations were not fully separate in this analysis, primarily because coho detection data alone were too sparse to properly estimate model parameters.
For clarity, we do not assume any other constraint on the model parameters, which allows us to assess the improvement resulting from these priors alone.
