Estimation of Parameters in Cox's Proportional Hazard Model: Comparisons between Evolutionary Algorithms and the Newton-Raphson Approach

The numerical Newton-Raphson approach used to optimize the partial likelihood function in Cox's proportional hazard model does not converge when there are collinearity problems in the design matrix, so we proposed the use of evolutionary algorithms where a comparison is made using both approaches, taking into account the capacity of the model to make good hazard predictions. The results indicate similar performance when the Newton---Raphson model has a good starter point; otherwise, the evolutionary algorithms generate suboptimal solutions.