Version: 1.3-3
Date: 2016-11-04
Title: Ecological Inference
Author: Gary King <king@harvard.edu>, Molly Roberts <molly.e.roberts@gmail.com>
Maintainer: James Honaker <zelig.zee@gmail.com>
Depends: R (≥ 2.5.0), eiPack
Imports: mvtnorm, msm, tmvtnorm, ellipse, plotrix, MASS, ucminf, cubature, mnormt, foreach, sp
Suggests: rgl
Description: Software accompanying Gary King's book: A Solution to the Ecological Inference Problem. (1997). Princeton University Press. ISBN 978-0691012407.
License: GPL-2 | GPL-3 [expanded from: GPL (≥ 2)]
URL: http://gking.harvard.edu/eiR, http://gking.harvard.edu/eicamera/kinroot.html
NeedsCompilation: no
Packaged: 2016-11-04 22:19:49 UTC; tercer
Repository: CRAN
Date/Publication: 2016-11-05 00:49:10

Sample Dataset

Description

A description for this dataset

Usage

RxCdata

Format

A data frame containing 60 observations.

Source

Source

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press.

Computes Analytical Bounds from Accounting Identity

Description

Returns analytical bounds from accounting identity on unknown table relationships beta_b, beta_w, from known, observed, table marginals, x, t (and sample size n).

Usage

bounds1(x, t, n)

Arguments

x

vector of characteristics, e.g. percentage of blacks in each district

t

vector of characteristics, e.g. percentage of people that voted in each district

n

size of each observation, e.g. number of voters in each district

Author(s)

Gary King <<email: king@harvard.edu>> and Molly Roberts <<email: molly.e.roberts@gmail.com>>

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press.

Examples

	data(census1910)
	output<-bounds1(x=census1910$x, t=census1910$t, n=census1910$n)

Black Literacy in 1910

Description

A dataset of aggregate literacy rates (t) and fraction of the population that is black (x), from the 1910 US Census. Each observation represents one county.

Usage

census1910

Format

A data frame containing 1030 observations.

Source

Gary King, 1997, "Replication data for: A Solution to the Ecological Inference Problem: Reconstructing Individual Behavior from Aggregate Data", http://hdl.handle.net/1902.1/LWMMKUTYXS UNF:3:DRWozWd89+vNLO7lY2AHbg== IQSS Dataverse Network [Distributor] V3 [Version]

References

Gary King. (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press. Section 13.2:241-5.

Robinson, William S. (1950). “Ecological Correlation and the Behavior of Individuals.” American Sociological Review 15:351-357.

Ecological Inference Estimation

Description

ei is the main command in the package EI. It gives observation-level estimates (and various related statistics) of \beta_i^b and \beta_i^w given variables T_i and X_i (i=1,...,n) in this accounting identity: T_i=\beta_i^b*X_i + \beta_i^w*(1-X_i). Results are stored in an ei object, that can be read with summary() or eiread() and graphed in plot().

Usage

ei(formula, total = NULL, Zb = 1, Zw = 1, id = NA, data =NA, erho = 0.5, 
esigma = 0.5, ebeta = 0.5, ealphab = NA, ealphaw = NA, truth = NA, 
simulate = TRUE, covariate = NULL, lambda1 = 4, lambda2 = 2, 
covariate.prior.list = NULL, tune.list = NULL, start.list = NULL, 
sample = 1000, thin = 1, burnin = 1000, verbose = 0, ret.beta = "r", 
ret.mcmc = TRUE, usrfun = NULL)

Arguments

formula

A formula of the form t ~x in the 2x2 case and cbind(col1,col2,...) ~ cbind(row1,row2,...) in the RxC case.

total

‘total’ is the name of the variable in the dataset that contains the number of individuals in each unit

Zb

p x k^b matrix of covariates or the name of covariates in the dataset

Zw

p x k^w matrix of covariates or the name of covariates in the dataset

id

‘id’ is the nae of the variable in the dataset that identifies the precinct. Used for ‘movie’ and ‘movieD’ plot functions.

data

data frame that contains the variables that correspond to formula. If using covariates and data is specified, data should also contain Zb and Zw.

erho

The standard deviation of the normal prior on \phi_5 for the correlation. Default =0.5.

esigma

The standard deviation of an underlying normal distribution, from which a half normal is constructed as a prior for both \breve{\sigma}_b and \breve{\sigma}_w. Default = 0.5

ebeta

Standard deviation of the "flat normal" prior on \breve{B}^b and \breve{B}^w. The flat normal prior is uniform within the unit square and dropping outside the square according to the normal distribution. Set to zero for no prior. Setting to positive values probabilistically keeps the estimated mode within the unit square. Default=0.5

ealphab

cols(Zb) x 2 matrix of means (in the first column) and standard deviations (in the second) of an independent normal prior distribution on elements of \alpha^b. If you specify Zb, you should probably specify a prior, at least with mean zero and some variance (default is no prior). (See Equation 9.2, page 170, to interpret \alpha^b).

ealphaw

cols(Zw) x 2 matrix of means (in the first column) and standard deviations (in the second) of an independent normal prior distribution on elements of \alpha^w. If you specify Zw, you should probably specify a prior, at least with mean zero and some variance (default is no prior). (See Equation 9.2, page 170, to interpret \alpha^w).

truth

A length(t) x 2 matrix of the true values of the quantities of interest.

simulate

default = TRUE:see documentation in eiPack for options for RxC ei.

covariate

see documentation in eiPack for options for RxC ei.

lambda1

default = 4:see documentation in eiPack for options for RxC ei.

lambda2

default = 2:see documentation in eiPack for options for RxC ei.

covariate.prior.list

see documentation in eiPack for options for RxC ei.

tune.list

see documentation in eiPack for options for RxC ei.

start.list

see documentation in eiPack for options for RxC ei.

sample

default = 1000

thin

default = 1

burnin

default = 1000

verbose

default = 0:see documentation in eiPack for options for RxC ei.

ret.beta

default = "r": see documentation in eiPack for options for RxC ei.

ret.mcmc

default = TRUE: see documentation in eiPack for options for RxC ei.

usrfun

see documentation in eiPack for options for RxC ei.

Details

The EI algorithm is run using the ei command. A summary of the results can be seen graphically using plot(ei.object) or numerically using summary(ei.object). Quantities of interest can be calculated using eiread(ei.object).

Author(s)

Gary King <<email: king@harvard.edu>> and Molly Roberts <<email: molly.e.roberts@gmail.com>>

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press.

Examples

data(sample)
form <- t ~ x
dbuf <- ei(form,total="n",data=sample)
summary(dbuf)

Simulate EI Solution via Importance Sampling

Description

Simulate EI solution via importance sampling

Usage

ei.sim(ei.object)

Arguments

ei.object

ei object

Author(s)

Gary King <<email: king@harvard.edu>> and Molly Roberts <<email: molly.e.roberts@gmail.com>>

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press.

A Sample Dataset

Description

A description for this dataset

Usage

eiRxCsample

Format

A data frame containing 93 observations.

Source

Source

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press.

Quantities of Interest from Ecological Inference Estimation

Description

eiread is the command that pulls quantities of interest from the ei object. The command returns a list of quantities of interest requested by the user.

Usage

eiread(ei.object, ...)

Arguments

ei.object

An ei object from the function ei.

...

A list of quantities of interest for eiread() to return. See values below.

Value

betab

p x 1 point estimate of \beta_i^b based on its mean posterior. See section 8.2

betaw

p x 1 point estimate of \beta_i^w based on its mean posterior. See section 8.2

sbetab

p x 1 standard error for the estimate of \beta_i^b, based on the standard deviation of its posterior. See section 8.2

sbetaw

p x 1 standard error for the estimate of \beta_i^w, based on the standard deviation of its posterior. See section 8.2

phi

Maximum posterior estimates of the CML

psisims

Matrix of random simulations of \psi. See section 8.2

bounds

p x 4: bounds on \beta_i^b and \beta_i^w, lowerB ~ upperB ~ lowerW ~ upperW. See Chapter 5.

abounds

2 x 2: aggregate bounds rows:lower, upper; columns: betab, betaw. See Chapter 5.

aggs

Simulations of district-level quantities of interest \hat{B^b} and \hat{B^w}. See Section 8.3.

maggs

Point estimate of 2 district-level parameters, \hat{B^b} and \hat{B^w} based on the mean of aggs. See Section 8.3.

VCaggs

Variance matrix of 2 district-level parameters, \hat{B^b} and \hat{B^w}. See Section 8.3.

CI80b

p x 2: lower~upper 80\% confidence intervals for \beta_i^b. See section 8.2.

CI80w

p x 2: lower~upper 80\% confidence intervals for \beta_i^w. See section 8.2.

eaggbias

Regressions of estimated \beta_i^b and \beta_i^w on a constant term and X_i.

goodman

Goodman's Regression. See Section 3.1

Author(s)

Gary King <<email: king@harvard.edu>> and Molly Roberts <<email: molly.e.roberts@gmail.com>>

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press.

Examples

data(sample)
formula = t ~ x
dbuf <- ei(formula=formula, total="n",data=sample)
eiread(dbuf, "phi")
eiread(dbuf, "betab", "betaw")

Voter Transitions

Description

Aggregated data from 289 precincts in Fulton County, Georgia. The variable t represents the fraction voting in 1994 and x the fraction in 1992. Beta_b is then the fraction who vote in both elections, and Beta_w the fraction of nonvoters in 1992 who vote in the midterm election of 1994.

Usage

fultongen

Format

A data frame containing 289 observations.

Source

Gary King, 1997, "Replication data for: A Solution to the Ecological Inference Problem: Reconstructing Individual Behavior from Aggregate Data", http://hdl.handle.net/1902.1/LWMMKUTYXS UNF:3:DRWozWd89+vNLO7lY2AHbg== IQSS Dataverse Network [Distributor] V3 [Version]

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press. Section 13.1:235-41.

Turnout by Race in Louisiana

Description

The fraction of blacks registered voters (x) and fraction of voter turnout (t) in each Louisiana precinct, along with the true fraction of black turnout (tb) and non-black turnout (tw).

Usage

lavoteall

Format

A data frame containing 3262 observations.

Source

Gary King, 1997, "Replication data for: A Solution to the Ecological Inference Problem: Reconstructing Individual Behavior from Aggregate Data", http://hdl.handle.net/1902.1/LWMMKUTYXS UNF:3:DRWozWd89+vNLO7lY2AHbg== IQSS Dataverse Network [Distributor] V3 [Version]

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press. Section 1.4:22-4.

Voter Registration by Race in Southern States

Description

Aggregate voter registration and fraction black, in counties in Florida, Louisiana, North Carolina and South Carolina

Usage

matproii

Format

A data frame containing 268 observations.

Source

Gary King, 1997, "Replication data for: A Solution to the Ecological Inference Problem: Reconstructing Individual Behavior from Aggregate Data", http://hdl.handle.net/1902.1/LWMMKUTYXS UNF:3:DRWozWd89+vNLO7lY2AHbg== IQSS Dataverse Network [Distributor] V3 [Version]

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press. Chapter 10.

Nonminority Turnout in New Jersey

Description

A description for this dataset

Usage

nj

Format

A data frame containing 493 observations.

Source

Gary King, 1997, "Replication data for: A Solution to the Ecological Inference Problem: Reconstructing Individual Behavior from Aggregate Data", http://hdl.handle.net/1902.1/LWMMKUTYXS UNF:3:DRWozWd89+vNLO7lY2AHbg== IQSS Dataverse Network [Distributor] V3 [Version]

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press. Section 1.4:24-5.

Plotting Ecological Inference Estimates

Description

‘plot’ method for the class ‘ei’.

Usage

## S3 method for class 'ei'
plot(x, ...)

Arguments

x

An ei object from the function ei.

...

A list of options to return in graphs. See values below.

Details

Returns any of a set of possible graphical objects, mirroring those in the examples in King (1997). Graphical option lci is a logical value specifying the use of the Law of Conservation of Ink, where the implicit information in the data is represented through color gradients, i.e. the color of the line is a function of the length of the tomography line. This can be passed as an argument and is used for “tomogD” and “tomog” plots.

Value

tomogD

Tomography plot with the data only. See Figure 5.1, page 81.

tomog

Tomography plot with ML contours. See Figure 10.2, page 204.

tomogCI

Tomography plot with 80\% confidence intervals. Confidence intervals appear on the screen in red with the remainder of the tomography line in yellow. The confidence interval portion is also printed thicker than the rest of the line. See Figure 9.5, page 179.

tomogCI95

Tomography plot with 95\% confidence intervals. Confidence intervals appear on the screen in red with the remainder of the tomography line in yellow. The confidence interval portion is also printed thicker than the rest of the line. See Figure 9.5, page 179.

tomogE

Tomography plot with estimated mean posterior \beta_i^b and \beta_i^w points.

tomogP

Tomography plot with mean posterior contours.

betab

Density estimate (i.e., a smooth version of a histogram) of point estimates of \beta_i^b's with whiskers.

betaw

Density estimate (i.e., a smooth version of a histogram) of point estimates of \beta_i^w's with whiskers.

xt

Basic X_i by T_i scatterplot.

xtc

Basic X_i by T_i scatterplot with circles sized proportional to N_i.

xtfit

X_i by T_i plot with estimated E(T_i|X_i) and conditional 80\% confidence intervals. See Figure 10.3, page 206.

xtfitg

xtfit with Goodman's regression line superimposed.

estsims

All the simulated \beta_i^b's by all the simulated \beta_i^w's. The simulations should take roughly the same shape of the mean posterior contours, except for those sampled from outlier tomography lines.

boundXb

X_i by the bounds on \beta_i^b (each precinct appears as one vertical line), see the lines in the left graph in Figure 13.2, page 238.

boundXw

X_i by the bounds on \beta_i^w (each precinct appears as one vertical line), see the lines in the right graph in Figure 13.2, page 238.

truth

Compares truth to estimates at the district and precinct-level. Requires truth in the ei object. See Figures 10.4 (page 208) and 10.5 (page 210).

movieD

For each observation, one tomography plot appears with the line for the particular observation darkened. After the graph for each observation appears, the user can choose to view the next observation (hit return), jump to a specific observation number (type in the number and hit return), or stop (hit "s" and return).

movie

For each observation, one page of graphics appears with the posterior distribution of \beta_i^b and \beta_i^w and a plot of the simulated values of \beta_i^b and \beta_i^w from the tomography line. The user can choose to view the next observation (hit return), jump to a specific observation number (type in the number and hit return), or stop (hit “s" and return).

Author(s)

Gary King <<email: king@harvard.edu>> and Molly Roberts <<email: molly.e.roberts@gmail.com>>

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press.

Examples

data(sample)
formula = t ~ x
dbuf <- ei(formula=formula, total="n",data=sample)
plot(dbuf, "tomog")
plot(dbuf, "tomog", "betab", "betaw", "xtfit")

Sample Data for Black Votes

Description

A description for this dataset

Usage

sample

Format

A vector containing 141 observations.

Source

Source

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press.

Summarize Ecological Inference Estimates

Description

‘summary’ method for the class ‘ei’.

Usage

## S3 method for class 'ei'
summary(object, ...)

Arguments

object

An ei object from the function ei.

...

A list of options to return in graphs. See values below.

Author(s)

Gary King <<email: king@harvard.edu>> and Molly Roberts <<email: molly.e.roberts@gmail.com>>

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press.

Examples

data(sample)
formula = t ~ x
dbuf <- ei(formula=formula, total="n",data=sample)
print(summary(dbuf))

Plotting Ecological Inference Estimates with eiRxC information

Description

A tomography plot for an estimated Ecological Inference model in RxC data.

Usage

tomogRxC(formula, data, total=NULL, refine=100)

Arguments

formula

A formula of the form cbind(col1, col2,...)~cbind(row1,row2,...)

data

data that contains the data that corresponds to the formula

total

‘total’ is the name of the variable in the dataset that contains the number of individuals in each unit

refine

specifies the amount of refinement for the image. Higher numbers mean better resolution.

Author(s)

Gary King <<email: king@harvard.edu>> and Molly Roberts <<email: molly.e.roberts@gmail.com>>

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press.

Examples

data(RxCdata)
formula = cbind(turnout, noturnout) ~ cbind(white, black,hisp)
tomogRxC(formula, data=RxCdata)

Plotting 2x3 Ecological Inference Estimates in 3 dimensions

Description

A tomography plot in 3 dimensions for RxC Ecological Inference data and an estimated Ecological Inference model in RxC data.

Usage

tomogRxC3d(formula, data, total=NULL, lci=TRUE, estimates=FALSE, ci=FALSE, level=.95, 
	seed=1234, color=hcl(h=30,c=100,l=60), transparency=.75, light=FALSE, rotate=TRUE)

Arguments

formula

A formula of the form cbind(col1, col2,...)~cbind(row1,row2,...)

data

data that contains the data that corresponds to the formula

total

‘total’ is the name of the variable in the dataset that contains the number of individuals in each unit

lci

logical value specifying the use of the Law of Conservation of Ink, where the implicit information in the data is represented through color gradients, i.e. the color of the plane is a function of the area of the tomography plane.

estimates

logical value specifying whether the point estimates of \beta's are included for each observation on the tomography plot.

ci

logical value specifying whether the estimated confidence ellipse is included on the tomography plot.

level

numeric value from 0 to 1 specifying the significance level of the confidence ellipse; eg. .95 refers to 95% confidence ellipse.

seed

seed value for model estimation.

color

color of tomography planes if lci=F.

transparency

numeric value from 0 to 1 specifying transparency of tomography planes; 0 is entirely transparent.

light

logical value specifying whether lights should be included in the rgl interface. The inclusion of lights will create shadows in the plot that may distort colors.

rotate

logical value specifying whether the plot will rotate for 20 seconds.

Details

Requires rgl package and rgl viewer.

Author(s)

Gary King <<email: king@harvard.edu>>; Molly Roberts <<email: molly.e.roberts@gmail.com>>; Soledad Prillaman <<email: soledadartiz@fas.harvard.edu..

References

Gary King (1997). A Solution to the Ecological Inference Problem. Princeton: Princeton University Press.

Examples

data(RxCdata)
formula <- cbind(turnout, noturnout) ~ cbind(white, black, hisp)
tomogRxC3d(formula, RxCdata, total=NULL, lci=TRUE, estimates=TRUE, ci=TRUE, transparency=.5, 
	light=FALSE, rotate=FALSE)