The down selection process is useful for working quickly through a data set that may have a number of explanatory independent variables.  By starting with the "full" model that has all potential independent variables and reading the regression statistics correctly, we can determine if the current model is likely to be the best model for that regression type (i.e., linear, power, exponential, etc. ) or if a derivable model with fewer variables will work better.  This video demonstrates the process using a 3- independent variable data set and working through the linear, power and exponential regression model types.