Unfortunately our model fails both the normality assumption and the constant variance assumption. There are probably ways to subset the data that would ensure a more constant variance or normality but the full dataset doesn't
Unfortunately our model fails both the normality assumption and the constant variance assumption. There are probably ways to subset or transform the data that would ensure a more constant variance or normality but using the full data without any transformations leaves us with a model that fails some of the basic assumptions of a linear model.
## Discussion
At the outset our goal was to gain a greater understanding of how officials in Connecticut are compensated. The model may not have the best predictive capabilities and it also fails the constant variance and normality assumption tests, but it provided insight into CT employee compensation. The findings in the simple model using age and agency in the appendix were interesting because we made an extrapolation about being a 100 year old employee of the judicial branch and it gave a reasonable prediction. Ideally we would have had more time to explore subsets of the data, maybe it would have made sense to have a different model trained for each agency or union. Our resulting model showed that the
At the outset our goal was to gain a greater understanding of how officials in Connecticut are compensated. The model may not have the best predictive capabilities and it also fails the constant variance and normality assumption tests, but it provided insight into CT employee compensation. The findings in the simple model using age and agency in the appendix were interesting because we made an extrapolation about being a 100 year old employee of the judicial branch and it gave a reasonable prediction. Ideally we would have had more time to explore subsets of the data, maybe it would have made sense to have a different model trained for each agency or union. Our resulting model showed that there is definitely a correlation between the predictors we found to be influential and Connecticut state employee's annual salary but training a regression model to the given data is hard work.
