Page 3 - Regression Guideline for AMC
P. 3
Evalua6ng Mul6ple Regression Model FiRng Real Estate Appraising
• In real estate appraising, the es6mates derived from the group of comparable proper6es are applied to the subject property to predict the most likely sales price or value of the subject property based on its known characteris6cs. As such, regression analysis in this context is described as “predic6ve analy6cs”. We are concerned with how well the sta6s6cal model predicts the value of the subject property.
• Mul6ple regression is nothing more than a tool to help with appraising and underwri6ng. Its value is that it provides an objec6ve, mathema6cally derived and replicable es6mate of a property’s value. Used correctly and with good data, a regression model can provide accurate informa6on on how different property characteris6cs affect value within a given market.
• However, regression analysis is not foolproof. If used incorrectly or based on poor data from non-‐comparable proper6es, or data from only a small number of comparable proper6es, regression models can provide es6mates that are inaccurate with such wide error margins they do not help narrow down the subject property’s es6mated value to a useful range.
3
