Page 30 - Regression Guideline for AMC

Normal distribu6on of data
• Key ques:on 3: Prior to analyzing the property data was the distribu:on of last known sales price checked for normality and were sta:s:cal outliers included or excluded from the analysis and on what basis (i.e., +-­‐ 2.0 standard devia:ons)?
• So far, we have evaluated model fit in terms of one example property by examining the size of its residual or error term and 95% confidence interval. However, it is also important to get an assessment of how well the model fits the data for all of the examined proper6es. For most regression models, some of the cases analyzed will be predicted well (i.e., have small residuals) and some will be predicted less well (i.e., have large residuals). We are looking for a model that predicts well for most of the examined proper6es.
• The following graph shows the predicted and actual values of the 402 Orlando proper6es examined and does show that the model predicts well for the majority of the proper6es.
30