Bayesian Projects
Project III: The Association Between the Proportion of Deaths and the Vaccination Effort in Each State
Project III: Abstract
In this analysis we were interested in understanding the association between vaccination efforts and the proportion of deaths. Of interest was to carefully curate a subset of explanatory variables in the data and to implement a g-prior based Bayes factor routine. The selected variables were then used to implement two different models, one assuming constant variance in the error structure and the other considering heterogeneity. An out-of-sample prediction was ran where both models achieved relatively reasonable results. Posterior predictive checks and posterior predictive loss were then utilized to further understand each model. Model 2 seemed slightly more favored and we were able to conclude a significant relationship between the percentage of individuals with 2 doses in each state and the proportion of deaths.
Exploratory Data Analysis III
Examining the explanatory variables above we note particular linear relationships that are present when looked in association with the logit-transformed proportions.
Statistical Modeling and Analysis II
From the G-prior variable selection we proceed with our chosen subset of explanatory variables.
We then proceed to compare both models with their out-of-sample predictive performances with model 2 being more robust due to the added complexity of adding a variance structure to our model.