ABSTRACT
The statistical field of survival analysis focuses on the examination of time-to-event data. The proportional hazards (PH) model is the most widely used in multivariate survival analysis to examine the effects of various factors on survival time. The statistics, however, do not always support the PH models assumption of constant hazards. The power of the associated statistical tests is reduced when the PH assumption is broken, which leads to incorrect interpretation of the estimation results. The accelerated failure time (AFT) models, on the other hand, do not, like the PH model, assume constant hazards in the survival data. Additionally, the AFT models can be employed in place of the PH model if the constant hazards assumption violated. This study set out to examine how well the PH model and the AFT models performed when it came to identifying the proximate variables influencing under – five mortality from National Family Health Survey data in Uttar Pradesh. Three AFT models that were based on the Weibull, exponential, and log-normal distributions were the only ones discussed in this article. The research employing a graphical technique and a statistical test revealed that the NFHS-5 data set has non-proportional hazards. The log-normal AFT model was the most acceptable model among the ones studied, according to the Akaike information criterion (AIC).