Balancing The Data
Regression analysis is a process that begins with selecting a group of properties on which the regression analysis will be done. An important step in the Savvi Analytics regression process is gathering data on a set of properties that will be used to estimate how sales prices are affected by property characteristics. While it is important that the selected properties are similar to the property being appraised in terms of characteristics such as square footage, regression analysis require sufficient variation to adequately and accurately estimation the final model.
Click here to learn more.
if the selected properties reflect only a very narrow range of prices and characteristics, there is not enough variation to model to determine how a property characteristic affects price across different price points. Conversely, if the range of the selected properties is too broad, there is a higher chance that properties too dissimilar to the property being appraised will be included in the model results in over or underestimates if a given characteristic’s effect on sales price. The Savvi Analytics approach achieves the right balance comparability and range by selecting at least 200 properties that are 30% (plus or minus) of the appraised property’s square footage and within 20 years (plus or minus) of the appraised property’s age. Next, a time adjustment is applied to the sales prices of the properties to make them maximally comparable to the time of the current appraisal.