Page 21 - Regression Guideline for AMC

Understanding regression mechanics
• In this example, we can directly measure the error term/ residual for this property because we know from the MLS lis6ngs what its last sold price was ($339,100).
• Comparing this value to the predicted value ($448,384.90 ), we calculate the residual (actual -­‐ predicted) as being -­‐ $109,284. The regression model overes6mated the value of this property by over $100,000 or, expressed as a percentage of the known property sale price, the model overes6mated by 32% of the property value ((109,284/339,100)*100).
• This is probably not an acceptable margin of error for an appraisal. What are the possible problems with this model that led to it inaccurately es6ma6ng the value of this property? How could we have spo.ed them and how might we correct them to get a more accurate model?
21