R/earth Package Review: The Free Vs The $15,000 Product

[Bert Craytor]

I finally got around to looking at the latest version of Stephen Milborrows R Language "earth" package. Last time I used it about 12 years ago it was terrible. Now I can say that it's pretty comparable to the $15,000 product MARS from Salford Systems.

Here is the link: https://www.PacificVista.net/?page_id=129

This also has a link to one of my public GitHub repositories with some MLS data from Pacifica (1606 MLS sales transactions from 1/30/2001 through 1/30/2018 I used for a retro appraisal, stripped of the PropertyID (MLS rules) and a ton of fields that aren't too important. Also stripped of sales transactions missing important fields.

Also, included the instructions for executing "earth" with the data. You will see it is pretty easy to use. --- Well anyway if you don't start trying to tweak the parameters ....)

I also dumped the MARS report with graphs and an Excel spreadsheet that compares the output of "earth" and MARS. The knots produced are very similar. The R2 value also pretty similar.

You would be well advised to get good at using earth ....

(I assume you know how to download R and set it up on your machine. How to install and load packages. .... It is not that difficult. You can Google "R Language". Bruce Hahn and others on this forum use R. Dell teaches courses in R. ...

 

[Red Flint]

R based Market Area Analysis


https://journal.r-project.org/archive/2017/RJ-2017-020/index.html


Model-based market area analysis for retail and service locations includes 1) theoretical considerations,

2) collection of empirical data regarding the locations and the customers, 3) data processing and, in

some cases, data transformation, 4) fitting of the used models and 5) the market area prediction itself.

While the first two aspects can not be offered by a statistical software, the presented package MCI

provides several functions to process the other mentioned steps by implementing the Huff Model and

the Multiplicative Competitive Interaction (MCI) Model into R.

 

[Bert Craytor]

R based Market Area Analysis


https://journal.r-project.org/archive/2017/RJ-2017-020/index.html


Model-based market area analysis for retail and service locations includes 1) theoretical considerations,

2) collection of empirical data regarding the locations and the customers, 3) data processing and, in

some cases, data transformation, 4) fitting of the used models and 5) the market area prediction itself.

While the first two aspects can not be offered by a statistical software, the presented package MCI

provides several functions to process the other mentioned steps by implementing the Huff Model and

the Multiplicative Competitive Interaction (MCI) Model into R.

Such models are interesting but do not work well in many areas around the Silicon Valley, where there are numerous constraints on where stores can be built or expanded. Invariably, they just tear down and/or expand existing retail sites. Along the coast, you have a laundry list of regulations, government employees and contractors dedicated to obstruction of anything they think would harm the environment, plus a host of local environmentalists.

We do get these exceptions that come along. Maybe low cost multi-family housing will be one. Where do you place new structures? Such techniques may be applicable here.

In any case, thanks for the link to R publications. I may be able to use it. Latex is not an obstacle (Latex is a language for producing documents commonly used in the science community.)