Bert,
I am just not sure how much I would go along with these things being “advanced.” I find much appraisal literature lacks even a fundamental understanding of arithmetic to the point that inverse operations are portrayed as separate “approaches” and “methods” that only varied in the report because of rounding error – replete with extensive explanations based on appraisal and economic theories attempting to “reconcile” the differences in the final reconciliation section of the report.
A second and related factor, for me, is recognizing how little we have to scratch surface of deep bodies of knowledge. Just because we have to figure out square footage of triangular areas, doesn’t mean we have to be full-fledged geometricians, reciting Euclid. Same thing with regression. We are just using a few mathematical tools that are also used by others.
Understanding these techniques is more a matter of simple vision and basic theory rather than profound knowledge. Maybe my viewpoint is tempered because when I first figure these things out, I had to do it with pencil and paper. It’s easy for anyone to understand if you pair two sales that you look at the change of price between them, and then change of square feet, and even just divide in your head to get something like $30/sf. If one does “regression” by hand, one sees that it is just summing all the prices and summing all the square feet (and some squaring and cross-multiplying, yada-yada), and its just comparing them all at once, instead of one pair at a time, and that it works better because the pairs don’t have to be “perfectly matched.” To me, the basic regression formula is just a better formula than the one used for perfectly matched pairs, because it works when the pairs are not perfectly matched (which they never are).
Steven,
1. The adjective "advanced" is in relation to appraiser knowledge. However, Multiple Adaptive Regression Splines, or more generally, "Generalized Additive Models" are actually one of the more "advanced" areas in statistics - even for statisticians - as many of the algorithms have only matured in the last decade. Some of the original works are "Generalized Additive Models" b T.J. Hastie & R.J. Tibshirani, 1990,(Stanford), "Classification & Regression Trees" by L. Breiman & J.H. Friedman, 1984( UC Berkeley & Stanford).
2. "Understanding the techniques" probably is not as important as understanding the nature of what they are applied to. If you don't believe me, see how far you get reading the two above books!
The best example is GLA, which is almost always the primary VARIABLE factor in contributory value (the major factor being base site value in California). A regression model may kick out a value of $300/sf. In fact, I have models of condos where the value is $1000/sf and high end homes with values of $800/sf. The "technique" does not understand what that means. However, the appraiser has to often have a very good, fundamental, in-depth understanding of the contributory value of various components of a property to put that kind of contributory value in its proper place - and possibly adjust it. This is a subject that takes some time to get into - and I won't try to do it here. Suffice it to say that in my opinion, the percentage of appraisers who could competently deal with this are those with good quality college degrees or who in any case could make it into Mensa or get at least 650 on the math portion of the SAT (and that doesn't guarantee anything of course ... because I know people that have advanced degrees, but have long since forgotten everything related to advanced math and science that they learned when they were young through disuse [very common ... I should know]).
The point here is that it is probably not profitable for a software company to come out with a product that supports regression to any degree. TRAINING would be a major expense. Would appraisers shell out the necessary $$$ for ADEQUATE training?
I am sure that current software products will slowly and incrementally be introduced as minor improvements in existing software ... but the appraisal software companies will be continually testing the market to see if what they are doing is profitable.
The other approach is that, that Snapwire is kind of taking... build a product (theirs is for commercial appraisal) that can, to a greater or lesser extent, work for other markets. That's because valuation sucks as a business. You've really got just one major group (Argus/SnapWire) that takes a majority of the commercial valuation business. Argus has actually has a near monopoly on DCF related products.
Other areas somewhat related to Valuation are far more profitable. For example, look at Pricing software - you see a much different story. (Noting that Pricing should have something to do with Valuation - but is in fact in a completely different ballbark.)
Bert Craytor, SRA
