Modeling survival response using a parametric approach in the presence of multicollinearity

Survival times with valid distributional assumptions are hard to treat when the assumption of no multicollinearity between predictors is violated. The model that evaluates the performance of survival functions provides an efficient tool to determine accurate systems. However, the standard methods addressing survival functions have some limitations in long-term evaluations in the presence of high multicollinearity with valid distributional assumptions. Therefore, a method that can treat survival outcomes on complicated functions is introduced which flexibly estimates the parameters. This method is developed by integrating a flexible parametric algorithm with partial least squares (PLS-FPM) to estimate continuous functions for extrapolating the survival modeling in long-term evaluations. The integrated method based on simulated data generated from Gompertz distributions is compared with standard partial least squares-Cox regression (PLS-CoxR). The analysis reflects the flexibility and accuracy of PLS-FPM in addressing survival time response. The method is also executed for four real data sets of breast cancer having survival responses. The regression coefficients of significant genes are estimated. The applicability in addressing several probability distributions as a parametric tool validates the flexibility of the method to a wider range of survival responses and evidence an effective instrument for the data-oriented evaluation of survival systems.