Notes

Now we can look at statistical modeling, which I believe is just as important for making business decisions as statistical analysis is.

For the modeling I used R and Rstudio
and with modeling comes some challenges and in this case it was dealing with correllation and the two different subsidiaries

To start off with getting rid of correllation I used correllation and regression tests for significance to remove the variables that equated to a 100% match on training data. Then to rid of bias based on injury year, I shuffled the index and seperated the data into the 75/25 rule for trainiing and testing data.

I then applied many different regression tests from ridge lasso and log to simple linear and full model linear and simple linear regression performed the best. I then performed model selcetion to get the most efficient model and wound up with the paramters consisting of injury year, high narcotic claim/yes, # of days off work, injury age, hand/arm injury, foot/leg injury, and injured worker hourly wage with high narcotic claim, hand/arm and foot/leg being factor variables