• Welcome to AppraisersForum.com, the premier online community for the discussion of real estate appraisal. Register a free account to be able to post and unlock additional forums and features.

These are the Morons Who are Running Climate with Our Government

gregb said:
Why don't you show and analysis for Bolinas, Fairfax or Stinson Beach?
Click to expand...

I've done Bolinas - back around 2005 or so. The sales there, back then were maybe once every couple of years and the property types quite different with little documentation and regression isn't going to get you anywhere. To be exact, it wasn't uncommon for house up for sale to sit on the market for more than two years - because the only buyer would be someone looking for a vacation home. You wre lucky to get an R2 of 0.40. At least back then. I don't know what the situation is now, I have to run regression when I get a chance. But, looking at the MLS, I don't see a problem AT ALL. We have 330 sales going back to 2019. I see patterns. MARS could definitely handle it.

Looking at Stinson Beach, I have 182 sales going back 2015, 430 sales going back 2000. It looks doable, but I am going venture it requires some extra work. I've done The Sea Ranch - and had to creates set of values or all houses related to distance to the beach, line of view and elevation. Doable - but extra work.

Fairfax I have done. I don't see a problem. We have over 500 sales going back to 2020.

In the worse of times, it is a sheer minority of homes the SF Bay Area. I don't deny such areas exist. In the SF Bay Area they are, in total, less than 1% of the market.

When I get a chance I will do a market analysis MARS regression on these areas.

But, the real point here is that YOU should be trying this yourself. You just need R and package earth.
 
RCA said:
But, the real point here is that YOU should be trying this yourself. You just need R and package earth.
Click to expand...
I’ll break this into three clean parts: R, what “earth” usually means, and how appraisers actually use this stuff in practice.



1.​


R is an open-source statistical programming language that’s very strong at:


  • Multiple regression (OLS, log-linear, nonlinear)
  • Hedonic price modeling
  • Diagnostics (multicollinearity, heteroskedasticity, outliers)
  • Graphing residuals and leverage points
  • Automating repeatable analysis

In appraisal terms, R is often used for:


  • Market condition (time) adjustments
  • Location or amenity modeling
  • Paired-sales alternatives when data is messy
  • Mass appraisal / hybrid studies
  • Litigation support (transparent, reproducible math)

R itself is just the engine — the real power comes from packages (libraries).



2. What appraisers usually mean by​


Short answer:


There is no mainstream appraisal or R package officially called “earth” that is known as “Package Earth” in appraisal practice.
Click to expand...

But there are two very common things people are usually referring to when they say this.



A)​


This is probably what you’re hearing — just said imprecisely.


In R, earth is a real package, and it stands for:


Enhanced Adaptive Regression Splines (MARS)
Click to expand...

What it does:


  • Builds nonlinear regression models
  • Automatically detects breakpoints and interactions
  • Allows curves instead of straight lines
  • Handles complex market behavior better than simple OLS

Why appraisers care:


  • Real estate markets are not linear
  • Price response to size, age, location often changes at thresholds
  • MARS can model:
    • diminishing returns to GLA
    • age effects that flatten out
    • neighborhood boundary effects

⚠️ Important caution
In lending appraisals, especially conventional SFR work:


  • earth models can be hard to explain to underwriters
  • USPAP requires explainability
  • MARS is better suited to:
    • market studies
    • litigation support
    • complex or atypical property analysis
    • mass appraisal environments

Many residential reviewers will reject a model they can’t easily follow, even if it’s statistically superior.



B) Confusion with​


Sometimes “package earth” is sloppy shorthand for spatial or geospatial tools, such as:


  • GIS-based regression
  • Latitude/longitude modeling
  • Distance-to variables (coast, CBD, freeway, hazards)
  • Google Earth–derived location attributes

In that case, they may mean:


  • R spatial packages (e.g., for mapping or distance modeling)
  • External GIS + regression workflow
  • Location intelligence layered into valuation

This is much more common in litigation or environmental proximity cases (airport noise, coastline, flood zones, Superfund sites — which you’ve dealt with).



3. How this fits into​


Here’s the honest appraisal-world breakdown:


Typical residential lending appraisal​


  • Simple OLS regression (or paired sales)
  • Possibly done in:
    • Excel
    • SPSS
    • R (basic lm())
  • Heavy emphasis on explanation, not model elegance

Advanced / litigation / consulting appraisal​


  • R is common
  • Packages like earth may be used
  • Models are:
    • documented
    • tested
    • defensible under cross-examination
  • Regression supports opinions, not replaces judgment

Mass appraisal / AVM / assessor work​


  • earth (MARS), random forests, boosted trees all common
  • USPAP burden is different
  • Explainability handled statistically rather than narratively


Bottom line​


  • ✅ R = powerful statistical language used for regression in appraisal
  • ✅ earth is an R package, not a generic “earth package”
  • ⚠️ It models nonlinear relationships, which is powerful but harder to defend in lending appraisals
  • ❌ There is no standalone appraisal product called “Package Earth”
 
gregb said:
I’ll break this into three clean parts: R, what “earth” usually means, and how appraisers actually use this stuff in practice.



1.​


R is an open-source statistical programming language that’s very strong at:


  • Multiple regression (OLS, log-linear, nonlinear)
  • Hedonic price modeling
  • Diagnostics (multicollinearity, heteroskedasticity, outliers)
  • Graphing residuals and leverage points
  • Automating repeatable analysis

In appraisal terms, R is often used for:


  • Market condition (time) adjustments
  • Location or amenity modeling
  • Paired-sales alternatives when data is messy
  • Mass appraisal / hybrid studies
  • Litigation support (transparent, reproducible math)

R itself is just the engine — the real power comes from packages (libraries).



2. What appraisers usually mean by​


Short answer:




But there are two very common things people are usually referring to when they say this.



A)​


This is probably what you’re hearing — just said imprecisely.


In R, earth is a real package, and it stands for:




What it does:


  • Builds nonlinear regression models
  • Automatically detects breakpoints and interactions
  • Allows curves instead of straight lines
  • Handles complex market behavior better than simple OLS

Why appraisers care:


  • Real estate markets are not linear
  • Price response to size, age, location often changes at thresholds
  • MARS can model:
    • diminishing returns to GLA
    • age effects that flatten out
    • neighborhood boundary effects

⚠️ Important caution
In lending appraisals, especially conventional SFR work:


  • earth models can be hard to explain to underwriters
  • USPAP requires explainability
  • MARS is better suited to:
    • market studies
    • litigation support
    • complex or atypical property analysis
    • mass appraisal environments

Many residential reviewers will reject a model they can’t easily follow, even if it’s statistically superior.



B) Confusion with​


Sometimes “package earth” is sloppy shorthand for spatial or geospatial tools, such as:


  • GIS-based regression
  • Latitude/longitude modeling
  • Distance-to variables (coast, CBD, freeway, hazards)
  • Google Earth–derived location attributes

In that case, they may mean:


  • R spatial packages (e.g., for mapping or distance modeling)
  • External GIS + regression workflow
  • Location intelligence layered into valuation

This is much more common in litigation or environmental proximity cases (airport noise, coastline, flood zones, Superfund sites — which you’ve dealt with).



3. How this fits into​


Here’s the honest appraisal-world breakdown:


Typical residential lending appraisal​


  • Simple OLS regression (or paired sales)
  • Possibly done in:
    • Excel
    • SPSS
    • R (basic lm())
  • Heavy emphasis on explanation, not model elegance

Advanced / litigation / consulting appraisal​


  • R is common
  • Packages like earth may be used
  • Models are:
    • documented
    • tested
    • defensible under cross-examination
  • Regression supports opinions, not replaces judgment

Mass appraisal / AVM / assessor work​


  • earth (MARS), random forests, boosted trees all common
  • USPAP burden is different
  • Explainability handled statistically rather than narratively


Bottom line​


  • ✅ R = powerful statistical language used for regression in appraisal
  • ✅ earth is an R package, not a generic “earth package”
  • ⚠️ It models nonlinear relationships, which is powerful but harder to defend in lending appraisals
  • ❌ There is no standalone appraisal product called “Package Earth”
Click to expand...

Well good for starters. It shows you are thinking for yourself - or you have a bad AI package.

1. The reason Stephen Milborrows MARS implementations is called "earth" is because Salford Systems had a trademark on the name MARS and wouldn't let anyone use it, Now that Salford Systems was bought out by MiniTab, it appears they are not so aggressively protecting the trademark. But that is why Milborrow called it "earth" - to make fun of the MARS trademark - and he wasn't the only one pissed by the trademark. The words behind the so-called acronym are just fortuitous.

2. So Stephen Milborrows code is indeed an official package of CRAN - the main R repository. And "R/earth package" is accurate. If you are in RStudio writing your MARS code, you will have to run "install.packages("earth") to download earth from CRAN. Then you can run it. Check Milborrows documentation on this.

3. To repeat: "The CRAN earth package is an implementation of Jerome Friedman's MARS algorithm, with some improvements.

--- I am an expert here - have used Salford Systems MARS and R/earth for well over 20 years.

And, it appears you have not yet succeeded in getting it to run. Not easy, because you will be running into various problems beginners have to learn about the hard way, such as overfitting and the importance of including certain variables. I'll let you find out on your own.

Multivariate adaptive regression spline - Wikipedia

en.wikipedia.org en.wikipedia.org
 
Last edited:

Revolving Green Door: Former Biden Officials Landed Jobs With Environmental NGOs After Funneling Money To Them​

teaser image


Shocker, we know.

As the Trump administration's Department of Energy moves to wipe out over $83 billion in "Green New Scam" loans and conditional commitments approved in the final months of the Biden administration, a new analysis reveals that not only did the rush to spend accelerate right after Biden's disastrous June 27, 2024 debate with now-President Donald Trump, senior Biden officials landed roles at organizations that received agency funding. In some cases, money was steered to NGOs that the officials worked for before joining the government - where they then returned following the cash bonanza.

... justice for all :ohmy: ... and no one goes to jail :ROFLMAO:
 
You must log in or register to reply here.
Find a Real Estate Appraiser - Enter Zip Code

Copyright © 2000-, AppraisersForum.com, All Rights Reserved
AppraisersForum.com is proudly hosted by the folks at AppraiserSites.com
Back
Top