Learn R Programming

spdep (version 0.6-15)

boston: Corrected Boston Housing Data

Description

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.

Usage

data(boston)

Arguments

Format

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

References

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.

Examples

Run this code
# NOT RUN {
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)
lm.morantest(hr0, nb2listw(boston.soi))
# }
# NOT RUN {
require(maptools)
boston.tr <- readShapePoly(system.file("etc/shapes/boston_tracts.shp",
  package="spdep")[1], ID="poltract",
  proj4string=CRS(paste("+proj=longlat +datum=NAD27 +no_defs +ellps=clrk66",
  "+nadgrids=@conus,@alaska,@ntv2_0.gsb,@ntv1_can.dat")))
boston_nb <- poly2nb(boston.tr)
# }
# NOT RUN {
gp1 <- errorsarlm(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, nb2listw(boston.soi), method="Matrix", 
 control=list(tol.opt = .Machine$double.eps^(1/4)))
summary(gp1)
gp2 <- lagsarlm(log(CMEDV) ~ CRIM + ZN + INDUS + CHAS + I(NOX^2) + I(RM^2)
 +  AGE + log(DIS) + log(RAD) + TAX + PTRATIO + B + log(LSTAT),
 data=boston.c, nb2listw(boston.soi), method="Matrix")
summary(gp2)
# }
# NOT RUN {
## Conversion table 1980/1970
# ICPSR_07913.zip
# 07913-0001-Data.txt
# http://dx.doi.org/10.3886/ICPSR07913.v1
# Provider: ICPSR
# Content: text/plain; charset="us-ascii"
# 
# TY  - DATA
# T1  - Census of Population and Housing 1980 [United States]:
# 1970-Pre 1980 Tract Relationships
# AU  - United States Department of Commerce. Bureau of the Census
# DO  - 10.3886/ICPSR07913.v1
# PY  - 1984-06-28
# UR  - http://dx.doi.org/10.3886/ICPSR07913.v1
# PB  - Inter-university Consortium for Political and Social Research
# (ICPSR) [distributor]
# ER  -
# widths <- c(ID=5L, FIPS70State=2L, FIPS70cty=3L, Tract70=6L, FIPS80State=2L,
#  FIPS80cty=3L, f1=7L, CTC=6L, f2=2L, intersect1=3L, intersect2=3L, name=30L)
# dta0 <- read.fwf("07913-0001-Data.txt", unname(widths),
#  col.names=names(widths), colClasses=rep("character", 12), as.is=TRUE)
# sub <- grep("25", dta0$FIPS80State)
# MA <- dta0[sub,]
## match against boston data set
# library(spdep)
# data(boston)
# bTR <- boston.c$TRACT
# x1 <- match(as.integer(MA$Tract70), bTR)
# BOSTON <- MA[!is.na(x1),]
## MA 1990 tracts
# library(rgdal)
# MAtr90 <- readOGR(".", "tr25_d90")
## counties in the BOSTON SMSA
## https://www.census.gov/population/metro/files/lists/historical/90nfips.txt
## 1123		Boston-Lawrence-Salem-Lowell-Brockton, MA NECMA
## 1123 25 009	  Essex County
## 1123 25 017	  Middlesex County
## 1123 25 021	  Norfolk County
## 1123 25 023	  Plymouth County
## 1123 25 025	  Suffolk County
# BOSTON_SMSA <- MAtr90[MAtr90$CO <!-- %in% c("009", "017", "021", "023", "025"),] -->
# proj4string(BOSTON_SMSA) <- CRS(paste("+proj=longlat +datum=NAD27 +no_defs",
#   "+ellps=clrk66 +nadgrids=@conus,@alaska,@ntv2_0.gsb,@ntv1_can.dat"))
# CTC4 <- substring(BOSTON$CTC, 1, 4)
# CTC4u <- unique(CTC4)
# TB_CTC4u <- match(BOSTON_SMSA$TRACTBASE, CTC4u)
## match 1980 tracts with 1990
# BOSTON_SMSA1 <- BOSTON_SMSA[!is.na(TB_CTC4u),]
## union Polygons objects with same 1970 tract code
#library(rgeos)
# BOSTON_SMSA2 <- gUnaryUnion(BOSTON_SMSA1,
#  id=as.character(BOSTON_SMSA1$TRACTBASE))
## reorder data set
# mm <- match(as.integer(as.character(row.names(BOSTON_SMSA2))), boston.c$TRACT)
# df <- boston.c[mm,]
# row.names(df) <- df$TRACT
# row.names(BOSTON_SMSA2) <- as.character(as.integer(row.names(BOSTON_SMSA2)))
## create SpatialPolygonsDataFrame
# BOSTON_SMSA3 <- SpatialPolygonsDataFrame(BOSTON_SMSA2,
#  data=data.frame(poltract=row.names(BOSTON_SMSA2),
#  row.names=row.names(BOSTON_SMSA2)))
# BOSTON_SMSA4 <- spCbind(BOSTON_SMSA3, df)
# mm1 <- match(boston.c$TRACT, row.names(BOSTON_SMSA4))
# BOSTON_SMSA5 <- BOSTON_SMSA4[mm1,]
#writeOGR(BOSTON_SMSA5, ".", "boston_tracts", driver="ESRI Shapefile",
# overwrite_layer=TRUE)
# moran.test(boston.c$CMEDV, nb2listw(boston.soi))
# moran.test(BOSTON_SMSA5$CMEDV, nb2listw(boston.soi))
# }

Run the code above in your browser using DataLab