hpiR

This package intends to simplify and standardize the creation of house price indexes in R. It also provides a framework for judging the quality of a given index by testing for predictive accuracy, volatility and revision. By providing these metrics various index methods (and estimators) can be accurately compared against each other.

While there are a (ever-increasing) variety of methods and models to use in house price index creation, this initial version (0.3.0) focuses on the two most common: repeat sales (transactions) and hedonic price. Base, robust and weighted estimators are provided when appropriate. Additionally, a new method using random forests and a post model interpretability method – partial dependence plots – is also used.

The package also includes a dataset of single family and townhome sales from the City of Seattle during January 2010 to December 2016 time period.

Please see the vignette for more information on using the package.

Also, please log issues or pull requests on this github page.

Installation

You can install hpiR from github with:

Install the released version from CRAN

  install.packages("hpiR")

Development version from GitHub:

  #install.packages("devtools")
  devtools::install_github("andykrause/hpiR")

Example

This is a basic example which shows you how to solve a common problem:

  library(hpiR)

  # Load prepared data
  data(ex_rtdata)

  # Create an index
  hpi <- rtIndex(trans_df = ex_rtdata,
                 estimator = 'robust',
                 log_dep = TRUE,
                 trim_model = TRUE,
                 smooth = TRUE)