Regression Analysis
I have just about finished Part I of the Regression Labs initiative. This includes a discussion of market modeling (which includes AVMs). It will also include a down-loadable regression application that runs in Excel (this application will run until the end of the year). It is designed to download data from your local MLS (Your MLS must download to a spreadsheet). I am looking for appraisers that have been using regression analysis and can give me some input. Be watching for this article on the Lab's site.
Labels: appraisals, AVM, Regression
posted by David Braun | Read entry and post comments (3 comments)
3 Comments:
Dave:
I finally read your paper. I like it, and agree with it. The models I have run so far have been pretty educational. I have tinkered with it plenty enough to see that you can almost make it work for any situation, but to get it right it can as much work as just doing the appraisal. It really is interesting, and in a just a few cases it proved to be off. I have run it on just about every condo, or other cookie cutter out there. The appraiser does need to be involved with data selection, and remove or augment things to get them correct. Case in point: the last detached condo I did has square footage all over the place. Once I cleaned it up ( I’ve measured each model many times) it ran pretty damn near perfect. It appraised it with a 96.4 rating, but still was about 5,000 high on a 260,000 appraised value. Not to shabby.
Here is the response I sent to an Email on this subject:
--Jim
I am not sure what you have read I do not see anything posted on the Labs web site yet except a short paragraph in the Blog. There will be a Part 1 of a three part article & a regression application that can be downloaded up on the site very soon.
The purpose of the 3 part paper is to educate appraisers with some facts instead of fiction. The Labs is looking for a good way for appraisers to pull from 20-80 sales from their MLS or Comps DB and analyze them for (1 a rough value, (2 justify adjustment amounts, and (3 rate the sales (by market fit & by adj ratios).
Part I is general, Part 2 is about regression (I will show how it is only really good at #1 above, and Part 3 will be the unveiling of a market modeling process that I believe will accomplish all 3 goals. Then based on the feedback Alamode will decide whether to make some of these tools available.
I am very excited about hearing from guys like you that have dealt with the limitations of auto market modeling methods. I do not want to answer too much about what type I am recommending that appraisers use, as it will be better understood after the three papers are published on the labs web site.
I will save this Email & call you next week to ask some questions of you.
Thanks
--David
The first part article on "Market Modeling" is now posted to the Labs website. It includes a regression application designed for appraisers (that runs in MS Excel) that you can download and use until the end of the year. Here is an excerpt:
A model is an explanation of an outcome. The usefulness of a model is its ability to predict some event. A model predicting when an egg in an incubator will hatch might consider the temperature and rotation of the eggs. A model addressing the number of wins and losses for a college football team might consider the average age, weight, time to run the 40, years played at college level, etc. Appraisers are concerned with predicting the sales price of a hypothetical sale within a particular real property sub-market. You have always been a market modeler and has been performing all of the steps and procedures of market modeling.
There will be a three part tutorial that you can watch to see how the regression model works.
Give me some feedback.
David
Post a Comment
<< Home