A Comparative Study Among Parametric, Semiparametric and Non-parametric Techniques Using Survival Data

This study offers a comparative analysis of parametric, semi-parametric, and non-parametric techniques of survival data analysis. It focuses on assessing the methods in terms of their efficiency in estimating survival probabilities and hazard functions. Alternate real and simulated datasets are then utilized to assess the relative strengths and weaknesses of different approaches with respect to efficiency, flexibility, and interpretability. The results indicate that the choice of technique is a factor of the underlying data characteristics, with parametric models working quite well under specific assumptions, semi parametric methods balancing between the structure and flexibility, while the nonparametric ones have the best performance in scenarios driven by the data. The AIC and BIC values determined from the model selection process demonstrated that the model finding fits the best under the given conditions as it has the smallest AIC and BIC scores. This is because AIC and BIC balance between goodness of fit and complexity in the model. For the Cox proportional hazard model results, most variables tested in the analysis showed very few statistically significant p- values. Results from Kaplan-Meier survival analysis strengthened the model by providing critical survival probability key stages over time, indicating the most occurrences beyond 60 months marked significant declines in survival rates. Hence, since the determined model com- bines predictive accuracy with interpret-ability, it should have as few but robust elements as would be meaningful in survival prediction without over-fitting. Therefore, the final model selection is based on the best compromise between explanatory power and statistical validity.