The boston.c data frame has 506 rows and 20 columns. It contains the Harrison and Rubinfeld (1978) data corrected for a few minor errors and augmented with the latitude and longitude of the observations. Gilley and Pace also point out that MEDV is censored, in that median values at or over USD 50,000 are set to USD 50,000. The original data set without the corrections is also included in package mlbench as BostonHousing. In addition, a matrix of tract point coordinates projected to UTM zone 19 is included as boston.utm, and a sphere of influence neighbours list as boston.soi.
This data frame contains the following columns:
TOWN: a factor with levels given by town names
TOWNNO: a numeric vector corresponding to TOWN
TRACT: a numeric vector of tract ID numbers
LON: a numeric vector of tract point longitudes in decimal degrees
LAT: a numeric vector of tract point latitudes in decimal degrees
MEDV: a numeric vector of median values of owner-occupied housing in USD 1000
CMEDV: a numeric vector of corrected median values of owner-occupied housing in USD 1000
CRIM: a numeric vector of per capita crime
ZN: a numeric vector of proportions of residential land zoned for lots over 25000 sq. ft per town (constant for all Boston tracts)
INDUS: a numeric vector of proportions of non-retail business acres per town (constant for all Boston tracts)
CHAS: a factor with levels 1 if tract borders Charles River; 0 otherwise
NOX: a numeric vector of nitric oxides concentration (parts per 10 million) per town
RM: a numeric vector of average numbers of rooms per dwelling
AGE: a numeric vector of proportions of owner-occupied units built prior to 1940
DIS: a numeric vector of weighted distances to five Boston employment centres
RAD: a numeric vector of an index of accessibility to radial highways per town (constant for all Boston tracts)
TAX: a numeric vector full-value property-tax rate per USD 10,000 per town (constant for all Boston tracts)
PTRATIO: a numeric vector of pupil-teacher ratios per town (constant for all Boston tracts)
B: a numeric vector of 1000*(Bk - 0.63)^2 where Bk is the
proportion of blacks
LSTAT: a numeric vector of percentage values of lower status population
Previously available from http://lib.stat.cmu.edu/datasets/boston_corrected.txt
Details of the creation of the tract GPKG file: tract boundaries for 1990 (formerly at: http://www.census.gov/geo/cob/bdy/tr/tr90shp/tr25_d90_shp.zip, counties in the BOSTON SMSA http://www.census.gov/population/metro/files/lists/historical/63mfips.txt); tract conversion table 1980/1970 (formerly at : https://www.icpsr.umich.edu/icpsrweb/ICPSR/studies/7913?q=07913&permit[0]=AVAILABLE, http://www.icpsr.umich.edu/cgi-bin/bob/zipcart2?path=ICPSR&study=7913&bundle=all&ds=1&dups=yes). The shapefile contains corrections and extra variables (tract 3592 is corrected to 3593; the extra columns are:
units: number of single family houses
cu5k: count of units under USD 5,000
c5_7_5: counts USD 5,000 to 7,500
C*_*: interval counts
co50k: count of units over USD 50,000
median: recomputed median values
BB: recomputed black population proportion
censored: whether censored or not
NOXID: NOX model zone ID
POP: tract population
Harrison, David, and Daniel L. Rubinfeld, Hedonic Housing Prices and the Demand for Clean Air, Journal of Environmental Economics and Management, Volume 5, (1978), 81-102. Original data.
Gilley, O.W., and R. Kelley Pace, On the Harrison and Rubinfeld Data, Journal of Environmental Economics and Management, 31 (1996),403-405. Provided corrections and examined censoring.
Pace, R. Kelley, and O.W. Gilley, Using the Spatial Configuration of the Data to Improve Estimation, Journal of the Real Estate Finance and Economics, 14 (1997), 333-340.
Bivand, Roger. Revisiting the Boston data set - Changing the units of observation affects estimated willingness to pay for clean air. REGION, v. 4, n. 1, p. 109-127, 2017. https://openjournals.wu.ac.at/ojs/index.php/region/article/view/107.
if (requireNamespace("spdep", quietly = TRUE)) {
data(boston)
hr0 <- lm(log(MEDV) ~ CRIM + ZN + INDUS + CHAS + I(NOX^2) + I(RM^2) +
AGE + log(DIS) + log(RAD) + TAX + PTRATIO + B + log(LSTAT), data = boston.c)
summary(hr0)
logLik(hr0)
gp0 <- lm(log(CMEDV) ~ CRIM + ZN + INDUS + CHAS + I(NOX^2) + I(RM^2) +
AGE + log(DIS) + log(RAD) + TAX + PTRATIO + B + log(LSTAT), data = boston.c)
summary(gp0)
logLik(gp0)
spdep::lm.morantest(hr0, spdep::nb2listw(boston.soi))
}
#>
#> Global Moran I for regression residuals
#>
#> data:
#> model: lm(formula = log(MEDV) ~ CRIM + ZN + INDUS + CHAS + I(NOX^2) +
#> I(RM^2) + AGE + log(DIS) + log(RAD) + TAX + PTRATIO + B + log(LSTAT),
#> data = boston.c)
#> weights: spdep::nb2listw(boston.soi)
#>
#> Moran I statistic standard deviate = 14.509, p-value < 2.2e-16
#> alternative hypothesis: greater
#> sample estimates:
#> Observed Moran I Expectation Variance
#> 0.4364296993 -0.0168870829 0.0009762383
#>
if (requireNamespace("sf", quietly = TRUE)) {
boston.tr <- sf::st_read(system.file("shapes/boston_tracts.gpkg",
package="spData")[1])
if (requireNamespace("spdep", quietly = TRUE)) {
boston_nb <- spdep::poly2nb(boston.tr)
}
}
#> Reading layer `boston_tracts' from data source
#> `/home/runner/work/_temp/Library/spData/shapes/boston_tracts.gpkg'
#> using driver `GPKG'
#> Simple feature collection with 506 features and 36 fields
#> Geometry type: POLYGON
#> Dimension: XY
#> Bounding box: xmin: -71.52311 ymin: 42.00305 xmax: -70.63823 ymax: 42.67307
#> Geodetic CRS: NAD27