Page 44 - Regression Guideline for AMC

MODEL STABILITY
• Models that are based on an adequate number of cases in the original data set have low shrinkage and are described as having high stability because their predic6ve accuracy changes very li.le when applied to a new set of data.
• Obtaining this many cases in rural or small market areas might be extremely challenging, however. In these markets it might be possible using expanded inclusion parameters such as extending the geographical area of the proper6es considered to obtain a larger sample database. We recommend at a minimum that the sample data base have at least 200 proper6es or about 20 proper6es per predictor if 10 property characteris6cs are analyzed.
• The Uniform Residen6al Appraisal Report (URAR) contains 14 different property characteris6cs, each of which could be included as a predictor in the regression model. However, modeling that many characteris6cs in a mul6ple regression model simultaneously would require a minimum of 280 proper6es (20 per predictor) and op6mally, 560 (40 per predictor).
44