Errors and mistakes can be “statistical.”  They can be modeling Blunders.  Or they can be of a third type:  misjudging what the data is showing you.

As we continue to grow the CAA (Community of Asset Analysts), we find that our body of useful algorithms and shared programs continues to deepen and widen.  The accuracy and usefulness of our valuation and risk functions continue to grow.  Our understanding of the speed, power, and reliability of our results continues to improve.

We have discovered that our competence in computation and “R” functionality must be built on a foundation of solid professional competence in the underlying economic, behavioral, and public policy issues.  In essence, no matter how clever the algorithm is, it requires the expert to:

  • Ask the right question (like the “scope of work”);
  • Select the right data classification solutions;
  • Apply the combination of predictives;
  • Deliver client decision needs.

My last peer-reviewed journal article (in Appraisal Institute’s The Appraisal Journal) was titled “Statistical Errors and Mistakes.”  What we know now is that the greatest mistakes come not from statistics itself, but from misfit models and myths.

This is the first in what is intended to be a series examining the nature and cause of the major mistakes made by analysts (and embedded in much current appraiser education).

It is hoped this will contribute to common understanding of how appraisers and asset analysts can deliver unbiased, high-reliability results, as well as additional needed products and services.

The best valuation results come from a merging of expert judgment and computer algorithms.  The professional expertise must be in modern data analytic methods.  The old “pick comps and make adjustments” approach simply embeds bias and uncertainty.

On the other hand, the new, modernized data-science approach does require the ability to identify and avoid the most common modeling myths.

A model is the process or algorithm that the analyst decides to apply to the problem-solution.  We will start with a brief list of the grossest and most destructive model myths.  I call them myths because they continue to be repeated in appraiser education, reinforced by misguided “common sense.”  Recall from your statistics class in high school:  Statistics is often not intuitive or common sensible!

The coming blogs of this series start with the inferential delusion.  This is the myth that somehow, some way, random sample inferential statistics can be forced upon the traditional appraisal ‘process.’  It cannot.

The second myth gets repeated and sold as a magical solution.  It is not.  This fallacy is that regression can be taken “as is” and applied to the appraisal problem.  But it cannot.  It takes modeling expertise by an expert in valuation.

This will be fun!