Page 35 - Regression Guideline for AMC
P. 35
R-‐Square
• As the preceding graphs illustrate, the value of the R-‐square sta6s6c reflects two related things: the strength of the rela6onship between the predictors and the dependent variable as well as the accuracy of the predicted values rela6ve to their actual values. It is possible for a model with high R-‐square to have a strong associa6on between the predictors and the dependent variable but s6ll also have a fair degree of inaccuracy. Conversely, a model with a low R-‐square can have good predic6ve accuracy but only a weak associa6on between the predictors and dependent variable.
• One way you can obtain a low R-‐square but have good predic6ve accuracy is when you have li.le varia6on on the dependent variable as would happen when only a small sample of proper6es was selected and their sales prices did not vary much. There is not much varia6on for the model to predict in this. This is one reason why it is important to have adequate varia6on not only for the property characteris6cs of the proper6es selected for the analysis but also for the range of sales prices reflected.
35
