Comparative Analysis of Bayesian Methods for Estimation of Locally-Invariant Extremes

Statistical methods are often fitting solutions to prediction and classification problems. Case studies have revealed that premature death is a priority problem in India and has often been rooted to malnutrition and birth-transmitted diseases. Aggregation of such medical and mortality records data helps to reveal important correlations between multiple factors affecting and aggravating the health conditions. The ground of research on the prediction of these diseases based on their severity has helped in opening up scopes of developing robust and swifter alternatives and taking affirmative actions to prevent further deaths. Bayesian methods have been proved to be more accurate in such case of spatial and temporal data points. This research explores the application of Local-Invariant and Independent features on various prolific algorithms like Bayesian Model Averaging, MHMCMC, RJ-MCMC and Bayesian Tail Regression. Further more likelihood estimators are used to predict the parameters and also measure the sampling accuracy. The proposed approach obtains a posterior probability of 91% for the simulated dataset from various regions involving several severity parameters.