Page 46 - Regression Guideline for AMC

MODEL STABILITY
• Only those characteris6cs that are sta6s6cally significant in the bivariate models are included in the final mul6variable model. We might have been able to reduce the number of predictors from 7 to 4 or 5 thereby reducing the number of proper6es needed for our final model by 20 to 40 proper6es per predictor excluded.
• The reason we can not simply use the results of the bivariate regressions for the final adjustments is that these models do not take into account the associa6ons between the predictors themselves. For example, proper6es with greater square footage are likely to have a larger number of bathrooms and bedrooms and to have larger acreage.
• The mul6ple regression model first es6mates then takes these associa6ons into account to obtain the final regression weights used for adjustments. Failure to account for the rela6onships among the predictors could lead to over-­‐ or under-­‐adjus6ng for their influence on property values and less accurate valua6on.
46