Consider another example in which we want to create a model for predicting the number of weekly operating errors in a manufacturing plant. The variables that have been measured include:

  • Number Of Weekly Operating Errors (the dependent variable)
  • Machine Downtime (%)
  • Number Of Workers

Let's start with a simple linear regression using machine downtime as the explanatory variable, which results in the equation:

# Of Weekly Errors = 50.49 − 3.82 × % Downtime

We can use a residual plot to help evaluate the fit of our simple regression to the data.

Line Fit Plot
A graph of percent downtime to number of weekly errors roughly fits a downward sloping trend line.
Residual Plot
A graph of percent downtime to residuals plots points with no discernible trend.

This residual plot does not show any discernible trend, so we can assume that for this simple regression, a linear relationship is the correct model.

However, suppose we want to turn this into a multiple regression by adding a second variable, to see if we can create a model that is a better fit for the data.