We are interested in investigating how to best estimate the number of physicians in a US county using various predictors. In the first part of our investigation, we wanted to know how we can improve our model if we only had log(TotalPop), LandArea, and IncPerCap as predictors. By finding the suitable transformations, we included log(TotalPop), log(LandArea), and IncPerCap in our model with log(Physicians) as the response.
In the second part, we examined the model with only TotalPop and Region. To build on our model, we selected other relevant predictors. Finally, we got log(TotalPop), Region, Bachelor, Poverty, and Pop65 as predictors with log(Physicians) as the response.