Page 32 - Regression Guideline for AMC
P. 32
R-‐Square
• Since it is not prac6cal to examine every case analyzed to determine how well the model fits the data, we must use summary sta6s6cs that tell us on average, how well the model predicts the dependent variable and how large the errors in predic6on are. We can then use these metrics to decide if the model’s accuracy allows us to use its es6mates to select comparable proper6es and appraise the value of a subject property.
• The most commonly used model fit sta6s6c in mul6ple regression analysis, by far, is called R-‐square or R2. R-‐square reflects the amount of varia6on in the dependent variable (sales price) accounted for by the set of predictors (property characteris6cs) in the regression model. While it is commonly shown in sta6s6cal print outputs as a decimal number (e.g., . 04, .50), it can easily be converted to a percentage by mul6plying by 100.
• As such, R-‐square can range from 0% meaning none of the varia6on in the dependent variable is “explained” by the predictors to 100% meaning all of the varia6on is explained.
32
