Content area
Full Text
So you are considering making appraisals with hedonic price modeling, namely using multiple regression analysis (MRA)? Fine! Unfortunately, we are currently inundated with charlatans and incompetents in the field. For example, a recent brochure from a gentleman who appears from his photo to be in his twenties, describes him as the "foremost authority in the country on the subject of regression analysis," and the author of the "first regression program ever written for real estate appraising!"
Apparently he is unaware that some of us were using MRA before he was born, or that a computer knows only two things-1 and 0-and has no idea what the numbers you are entering pertain to (in other words, no one needs a special regression program for real estate appraising. Even worse, he promises to educate you in applying regression to appraisal problems in a one-day course. And worse yet, a number of state appraiser boards have approved this guy's seminar for certification credits. So, first, let's have a quick little pop quiz, albeit an elementary one, to be quite sure that you know just what you are doing:
1. Do you know the difference between unadjusted R2 and adjusted R2, and do you know the consequences of using the unadjusted R2, and why you should never use the "regression" programs in Lotus 1-2-3 or your Hewlett-Packard?
2. Does your program tell you the tolerable R2, and do you know what its implications are?
3. Does it calculate the revised R2 so that you can see whether it is higher than the tolerable R2, which it must be for the regression to be valid for valuation purposes?
4. Do you know that "testing" a regression equation with part of its input data is meaningless when the regression is used for valuation or, in statistical terms, predicting a sale price, and that you need a...