Beyond predicting compensation with linear regression: find out the accuracy of the prediction and how close the data fit a line

Other posts on this blog have reviewed the basic notions of linear regression, using correlations between total compensation and various factors that determine it.  The calculation of a regression also tells how much of total compensation is predicted by each of the factors, such as years of law practice, practice area, size of department, industry, and so forth.

 

In one example, data from General Counsel Metrics shows that years out of law school only predicts about 50 percent of total compensation.

 

Additionally, software that calculates regressions can tell us how closely the data matches the regression line.  That number is known as the correlation of determination and the higher it reaches, the more the predictor attribute tracks total compensation.

 

Finally, what statisticians call a “t-test” tells how well the data conforms to a linear distribution. Many conclusions about data and inferences from it depend on the degree of linear correlation.