twang: Toolkit for Weighting and Analysis of Nonequivalent Groups

Description

Provides functions for propensity score estimating and weighting, nonresponse weighting, and diagnosis of the weights.

Subset of Alcohol and Other Drug treatment data

Description

A small subset of the data from McCaffrey et al. (2013).

Usage

data(AOD)

Format

A data frame with 600 observations on the following 10 variables.

treat

Treatment that each study subject received. Either community, metcbt5, or scy.

suf12

outcome variable, substance use frequency at 12 month follow-up

illact

covariate, illicit activities scale

crimjust

covariate, criminal justice involvement

subprob

covariate, substance use problem scale

subdep

covariate, substance use dependence scale

white

1 if non-Hispanic white, 0 otherwise

References

McCaffrey, DF, BA Griffin, D Almirall, ME Slaughter, R Ramchand and LF Burgette (2013). A tutorial on propensity score estimation for multiple treatments using generalized boosted models. Statistics in Medicine.

Calculate weighted balance statistics

Description

'bal.stat' compares the treatment and control subjects by means, standard deviations, effect size, and KS statistics

Usage

bal.stat(
  data,
  vars = NULL,
  treat.var,
  w.all,
  sampw,
  get.means = TRUE,
  get.ks = TRUE,
  na.action = "level",
  estimand,
  multinom,
  fillNAs = FALSE
)

Arguments

Details

'bal.stat' calls auxiliary functions for each variable and assembles the results in a table.

Value

'get.means' and 'get.ks' manipulate the inclusion of certain columns in the returned result.

References

Dan McCaffrey, G. Ridgeway, Andrew Morral (2004). "Propensity Score Estimation with Boosted Regression for Evaluating Adolescent Substance Abuse Treatment", *Psychological Methods* 9(4):403-425.

See Also

The example for [ps] contains an example of the use of [bal.table]

Compute the balance table.

Description

Extract the balance table from ps, dx.wts, and mnps objects

Usage

bal.table(
  x,
  digits = 3,
  collapse.to = c("pair", "covariate", "stop.method")[1],
  subset.var = NULL,
  subset.treat = NULL,
  subset.stop.method = NULL,
  es.cutoff = 0,
  ks.cutoff = 0,
  p.cutoff = 1,
  ks.p.cutoff = 1,
  timePeriods = NULL,
  ...
)

Arguments

Details

bal.table is a generic function for extracting balance tables from ps and dx.wts objects. These objects usually have several sets of candidate weights, one for an unweighted analysis and perhaps several stop.methods. bal.table will return a table for each set of weights combined into a list. Each list component will be named as given in the x, usually the name of the stop.method. The balance table labeled “unw” indicates the unweighted analysis.

Value

Returns a data frame containing the balance information.

• tx.mn The mean of the treatment group.

• tx.sd The standard deviation of the treatment group.

• ct.mn The mean of the control group.

• ct.sd The standard deviation of the control group.

• std.eff.sz The standardized effect size, (tx.mn-ct.mn)/tx.sd. If tx.sd is small or 0, the standardized effect size can be large or INF. Therefore, standardized effect sizes greater than 500 are set to NA.

• stat The t-statistic for numeric variables and the chi-square statistic for continuous variables.

• p The p-value for the test associated with stat ks The KS statistic.

• ks.pval The KS p-value computed using the analytic approximation, which does not necessarily work well with a lot of ties.

Boxplot for 'mnps' objects

Description

This function produces a collection of diagnostic plots for mnps objects.

Usage

## S3 method for class 'mnps'
boxplot(
  x,
  stop.method = NULL,
  color = TRUE,
  figureRows = NULL,
  singlePlot = NULL,
  multiPage = FALSE,
  time = NULL,
  print = TRUE,
  ...
)

Arguments

Details

This function produces lattice-style graphics of diagnostic plots.

References

Dan McCaffrey, G. Ridgeway, Andrew Morral (2004). "Propensity Score Estimation with Boosted Regression for Evaluating Adolescent Substance Abuse Treatment", *Psychological Methods* 9(4):403-425.

See Also

[mnps]

Boxplot for 'ps' objects

Description

This function produces a collection of diagnostic plots for ps objects.

Usage

## S3 method for class 'ps'
boxplot(x, subset = NULL, color = TRUE, time = NULL, ...)

Arguments

Details

This function produces lattice-style graphics of diagnostic plots.

References

Dan McCaffrey, G. Ridgeway, Andrew Morral (2004). "Propensity Score Estimation with Boosted Regression for Evaluating Adolescent Substance Abuse Treatment", *Psychological Methods* 9(4):403-425.

See Also

[ps]

Diagnosis of weights

Description

desc.wts assesses the quality of a set of weights on balancing a treatment and control group.

Usage

desc.wts(data, 
         w, 
         sampw = sampw,
         vars = NULL, 
         treat.var, 
         tp, 
         na.action = "level",
         perm.test.iters=0,
         verbose=TRUE,
         alerts.stack,
         estimand, multinom = FALSE, fillNAs = FALSE)

Arguments

Details

desc.wts calls bal.stat to assess covariate balance. If perm.test.iters>0 it will call bal.stat multiple times to compute Monte Carlo p-values for the KS statistics and the maximum KS statistic. It assembles the results into a list object, which usually becomes the desc component of ps objects that ps returns.

Value

See the description of the desc component of the ps object that ps returns

See Also

ps

Display plots

Description

Display plots

Usage

displayPlots(ptList, figureRows, singlePlot, multiPage, bxpt = FALSE)

Arguments

Compute diagnostics assessing covariates balance.

Description

dx.wts takes a ps object or a set of propensity scores and computes diagnostics assessing covariates balance.

Usage

dx.wts(
  x,
  data,
  estimand,
  vars = NULL,
  treat.var,
  x.as.weights = TRUE,
  sampw = NULL,
  perm.test.iters = 0
)

Arguments

Details

Creates a balance table that compares unweighted and weighted means and standard deviations, computes effect sizes, and KS statistics to assess the ability of the propensity scores to balance the treatment and control groups.

Value

Returns a list containing

• treat The vector of 0/1 treatment assignment indicators.

See Also

ps

US Sustaining Effects study

Description

A subset of the mathematics scores from the U.S. Sustaining Effects Study. The subset consists of information on 1721 students from 60 schools. This dataset is available in the mlmRev package.

Usage

data(egsingle)

Format

A data frame with 7230 observations on the following 12 variables.

schoolid

a factor of school identifiers

childid

a factor of student identifiers

year

a numeric vector indicating the year of the test

grade

a numeric vector indicating the student's grade

math

a numeric vector of test scores on the IRT scale score metric

retained

a factor with levels 0 1 indicating if the student has been retained in a grade.

female

a factor with levels Female Male

black

a factor with levels 0 1 indicating if the student is Black

hispanic

a factor with levels 0 1 indicating if the student is Hispanic

size

a numeric vector indicating the number of students enrolled in the school

lowinc

a numeric vector giving the percentage of low-income students in the school

mobility

a numeric vector

Source

Reproduced from themlmRev package for use in the section on nonresponse weighting in the twang package vignette. These data are distributed with the HLM software package (Bryk, Raudenbush, and Congdon, 1996). Conversion to the R format is described in Doran and Lockwood (2006).

References

Doran, H.C. and J.R. Lockwood (2006). “Fitting value-added models in R,” Journal of Educational and Behavioral Statistics, 31(1)

Extract propensity score weights.

Description

Extracts propensity score weights from a ps or mnps object.

Usage

get.weights(ps1, stop.method = NULL, estimand = NULL, withSampW = TRUE)

Arguments

Details

Weights for ATT are 1 for the treatment cases and p/(1-p) for the control cases. Weights for ATE are 1/p for the treatment cases and 1/(1-p) for the control cases.

Value

Returns a vector of weights.

See Also

ps, mnps

Get numerators to stabilize propensity score weights for 'iptw' fits.

Description

Forms numerators to stabilize weights for an iptw object.

Usage

get.weights.num(iptw, fitList)

Arguments

Value

Returns numerator of stabilized weights to be used in conjunction with 'get.weights.unstab'

See Also

[iptw]

Extract unstabilized propensity score weights for 'iptw' fits

Description

Extracts propensity score weights from an 'iptw' or 'mniptw' object.

Usage

get.weights.unstab(x, stop.method = NULL, withSampW = TRUE)

Arguments

Details

Weights are the reciprocal of the product of the probability of receiving the treatment received.

Value

Returns a data.frame of weights.

See Also

[iptw]

Inverse probability of treatment weighting for marginal structural models.

Description

iptw calculates propensity scores for sequential treatments using gradient boosted logistic regression and diagnoses the resulting propensity scores using a variety of methods

Usage

iptw(
  formula,
  data,
  timeInvariant = NULL,
  cumulative = TRUE,
  timeIndicators = NULL,
  ID = NULL,
  priorTreatment = TRUE,
  n.trees = 10000,
  interaction.depth = 3,
  shrinkage = 0.01,
  bag.fraction = 1,
  n.minobsinnode = 10,
  perm.test.iters = 0,
  print.level = 2,
  verbose = TRUE,
  stop.method = c("es.max"),
  sampw = NULL,
  version = "gbm",
  ks.exact = NULL,
  n.keep = 1,
  n.grid = 25,
  ...
)

Arguments

Details

For user more comfortable with the options of xgboost::xgboost(), the options for iptw controlling the behavior of the gradient boosting algorithm can be specified using the xgboost naming scheme. This includes nrounds, max_depth, eta, and subsample. In addition, the list of parameters passed to xgboost can be specified with params.

Value

Returns an object of class iptw, a list containing

psList

A list of ps objects with length equal to the number of time periods.

estimand

The specified estimand.

stop.methods

The stopping rules used to optimize iptw balance.

nFits

The number of ps objects (i.e., the number of distinct time points).

uniqueTimes

The unique times in the specified model.

See Also

ps, mnps, gbm, xgboost, plot, bal.table

Example data for iptw function (long version)

Description

These data are simulated to demonstrate the iptw function in the "long" data format.

Usage

data(lindner)

Format

A list with a covariate matrix and outcomes.

covariates

Time-invariant covariates are gender and age. The time-varying covariate is use. The reatment indicator is given by tx. An individual level identifier is given in ID, and the time period is time.

outcome

Vector of post-treatment outcomes.

Example data for iptw function (wide version)

Description

These data are simulated to demonstrate the iptw function in the "wide" data format.

Usage

data(lindner)

Format

A list with a covariate matrix and outcomes.

gender

Gender.

age

Age.

use0

Baseline substance use

.

use1

Use following first time period treatment.

use2

Use following second time period treatment.

tx1

Treatment indicator (first time period).

tx2

Treatment indicator (second time period).

tx3

Treatment indicator (third time period).

covariates

Time-invariant covariates are gender and age. The time-varying covariate is use. The reatment indicator is given by tx. An individual level identifier is given in ID, and the time period is time.

outcome

Post-treatment outcomes.

Lalonde's National Supported Work Demonstration data

Description

One of the datasets used by Dehejia and Wahba in their paper "Causal Effects in Non-Experimental Studies: Reevaluating the Evaluation of Training Programs." Also used as an example dataset in the MatchIt package.

Usage

data(lalonde)

Format

A data frame with 614 observations on the following 10 variables.

treat

1 if treated in the National Supported Work Demonstration, 0 if from the Current Population Survey

age

age

educ

years of education

black

1 if black, 0 otherwise

hispan

1 if Hispanic, 0 otherwise

married

1 if married, 0 otherwise

nodegree

1 if no degree, 0 otherwise

re74

earnings in 1974 (pretreatment)

re75

earnings in 1975 (pretreatment)

re78

earnings in 1978 (outcome)

Source

http://www.columbia.edu/~rd247/nswdata.html http://cran.r-project.org/src/contrib/Descriptions/MatchIt.html

References

Lalonde, R. (1986). Evaluating the econometric evaluations of training programs with experimental data. American Economic Review 76: 604-620.

Dehejia, R.H. and Wahba, S. (1999). Causal Effects in Nonexperimental Studies: Re-Evaluating the Evaluation of Training Programs. Journal of the American Statistical Association 94: 1053-1062.

Lindner Center data on 996 PCI patients analyzed by Kereiakes et al. (2000)

Description

These data are adapted from the lindner dataset in the USPS package. The description comes from that package, except for the variable sixMonthSurvive, which is a recode of lifepres

Data from an observational study of 996 patients receiving an initial Percutaneous Coronary Intervention (PCI) at Ohio Heart Health, Christ Hospital, Cincinnati in 1997 and followed for at least 6 months by the staff of the Lindner Center. The patients thought to be more severely diseased were assigned to treatment with abciximab (an expensive, high-molecular-weight IIb/IIIa cascade blocker); in fact, only 298 (29.9 percent) of patients received usual-care-alone with their initial PCI.

Usage

data(lindner)

Format

A data frame of 10 variables collected on 996 patients; no NAs.

lifepres

Mean life years preserved due to survival for at least 6 months following PCI; numeric value of either 11.4 or 0.

cardbill

Cardiac related costs incurred within 6 months of patient's initial PCI; numeric value in 1998 dollars; costs were truncated by death for the 26 patients with lifepres == 0.

abcix

Numeric treatment selection indicator; 0 implies usual PCI care alone; 1 implies usual PCI care deliberately augmented by either planned or rescue treatment with abciximab.

stent

Coronary stent deployment; numeric, with 1 meaning YES and 0 meaning NO.

height

Height in centimeters; numeric integer from 108 to 196.

female

Female gender; numeric, with 1 meaning YES and 0 meaning NO.

diabetic

Diabetes mellitus diagnosis; numeric, with 1 meaning YES and 0 meaning NO.

acutemi

Acute myocardial infarction within the previous 7 days; numeric, with 1 meaning YES and 0 meaning NO.

ejecfrac

Left ejection fraction; numeric value from 0 percent to 90 percent.

ves1proc

Number of vessels involved in the patient's initial PCI procedure; numeric integer from 0 to 5.

sixMonthSurvive

Survival at six months — a recoded version of lifepres.

References

Kereiakes DJ, Obenchain RL, Barber BL, et al. Abciximab provides cost effective survival advantage in high volume interventional practice. Am Heart J 2000; 140: 603-610.

Obenchain RL. (2009) USPSinR.pdf ../R_HOME/library/USPS 40 pages.

Extract table of means from an'mnps' object

Description

Extracts table of means from an mnps object.

Usage

means.table(mnps, stop.method = 1, includeSD = FALSE, digits = NULL)

Arguments

Details

Displays a table with weighted and unweighted means and standardized effect sizes, and – if requested – standard deviations.

Value

'A table of means, standardized effect sizes, and perhaps standard deviations, by treatment group.

See Also

[mnps]

Example data for iptw function (long version, more than two treatments).

Description

These data are simulated to demonstrate the iptw function in the "long" data format.

Usage

data(lindner)

Format

A list with a covariate matrix and outcomes.

covariates

Time-invariant covariates are gender and age. The time-varying covariate is use. The reatment indicator is given by tx. An individual level identifier is given in ID, and the time period is time.

outcome

Vector of post-treatment outcomes.

Example data for iptw function (wide version, more than two treatments)

Description

These data are simulated to demonstrate the iptw function in the "wide" data format.

Usage

data(lindner)

Format

A list with a covariate matrix and outcomes.

gender

Gender.

age

Age.

use0

Baseline substance use

.

use1

Use following first time period treatment.

use2

Use following second time period treatment.

tx1

Treatment indicator (first time period).

tx2

Treatment indicator (second time period).

tx3

Treatment indicator (third time period).

covariates

Time-invariant covariates are gender and age. The time-varying covariate is use. The reatment indicator is given by tx. An individual level identifier is given in ID, and the time period is time.

outcome

Post-treatment outcomes.

Propensity score estimation for multiple treatments

Description

mnps calculates propensity scores for more than two treatment groups using gradient boosted logistic regression, and diagnoses the resulting propensity scores using a variety of methods.

Usage

mnps(
  formula,
  data,
  n.trees = 10000,
  interaction.depth = 3,
  shrinkage = 0.01,
  bag.fraction = 1,
  n.minobsinnode = 10,
  perm.test.iters = 0,
  print.level = 2,
  verbose = TRUE,
  estimand = "ATE",
  stop.method = c("es.max"),
  sampw = NULL,
  version = "gbm",
  ks.exact = NULL,
  n.keep = 1,
  n.grid = 25,
  treatATT = NULL,
  ...
)

Arguments

Details

For user more comfortable with the options of xgboost::xgboost(), the options for mnps controlling the behavior of the gradient boosting algorithm can be specified using the xgboost naming scheme. This includes nrounds, max_depth, eta, and subsample. In addition, the list of parameters passed to xgboost can be specified with params.

Note that unlike earlier versions of twang, the plotting functions are no longer included in the mnps function. See plot for details of the plots.

Value

Returns an object of class mnps, which consists of the following.

psList

A list of ps objects with length equal to the number of time periods.

nFits

The number of ps objects (i.e., the number of distinct time points).

estimand

The specified estimand.

treatATT

For ATT fits, the treatment category that is considered "the treated".

treatLev

The levels of the treatment variable.

levExceptTreatAtt

The levels of the treatment variable, excluding the treatATT level.

data

The data used to fit the model.

treatVar

The vector of treatment indicators.

stopMethods

The stopping rules specified in the call to mnps.

sampw

Sampling weights provided to mnps, if any.

Author(s)

Lane Burgette '<burgette@rand.org>', Beth Ann Griffin '<bethg@rand.org>', Dan Mc- Caffrey '<danielm@rand.org>'

References

Dan McCaffrey, G. Ridgeway, Andrew Morral (2004). "Propensity Score Estimation with Boosted Regression for Evaluating Adolescent Substance Abuse Treatment", *Psychological Methods* 9(4):403-425.

See Also

ps, gbm, xgboost, plot, bal.table

Plot dxwts

Description

Plot dxwts

Usage

## S3 method for class 'dxwts'
plot(x, plots = "es", ...)

Arguments

Plots for iptw objects

Description

This function produces a collection of diagnostic plots for iptw objects.

Usage

## S3 method for class 'iptw'
plot(
  x,
  plots = "optimize",
  subset = NULL,
  color = TRUE,
  timePeriods = NULL,
  multiPage = FALSE,
  figureRows = NULL,
  hline = c(0.1, 0.5, 0.8),
  ...
)

Arguments

Details

This function produces lattice-style graphics of diagnostic plots.

References

Dan McCaffrey, G. Ridgeway, Andrew Morral (2004). "Propensity Score Estimation with Boosted Regression for Evaluating Adolescent Substance Abuse Treatment", Psychological Methods 9(4):403-425.

See Also

iptw

Plot mniptw

Description

Plot mniptw

Usage

## S3 method for class 'mniptw'
plot(
  x,
  plots = "optimize",
  pairwiseMax = TRUE,
  figureRows = NULL,
  color = TRUE,
  subset = NULL,
  treatments = NULL,
  singlePlot = NULL,
  multiPage = FALSE,
  timePeriods = NULL,
  hline = c(0.1, 0.5, 0.8),
  ...
)

Arguments

Plots for mnps objects

Description

This function produces a collection of diagnostic plots for mnps objects.

Usage

## S3 method for class 'mnps'
plot(
  x,
  plots = "optimize",
  pairwiseMax = TRUE,
  figureRows = NULL,
  color = TRUE,
  subset = NULL,
  treatments = NULL,
  singlePlot = NULL,
  multiPage = FALSE,
  time = NULL,
  print = TRUE,
  hline = c(0.1, 0.5, 0.8),
  ...
)

Arguments

Details

This function produces lattice-style graphics of diagnostic plots.

References

Dan McCaffrey, G. Ridgeway, Andrew Morral (2004). "Propensity Score Estimation with Boosted Regression for Evaluating Adolescent Substance Abuse Treatment", Psychological Methods 9(4):403-425.

See Also

mnps

Plots for ps objects

Description

This function produces a collection of diagnostic plots for ps objects.

Usage

## S3 method for class 'ps'
plot(x, plots = "optimize", subset = NULL, color = TRUE, ...)

Arguments

Details

This function produces lattice-style graphics of diagnostic plots.

References

Dan McCaffrey, G. Ridgeway, Andrew Morral (2004). "Propensity Score Estimation with Boosted Regression for Evaluating Adolescent Substance Abuse Treatment", Psychological Methods 9(4):403-425.

See Also

ps

Default print statement for dxwts class

Description

Default print statement for dxwts class

Usage

## S3 method for class 'dxwts'
print(x, ...)

Arguments

Default print statement for iptw class

Description

Default print statement for iptw class

Usage

## S3 method for class 'iptw'
print(x, ...)

Arguments

Default print statement for mniptw class

Description

Default print statement for mniptw class

Usage

## S3 method for class 'mniptw'
print(x, ...)

Arguments

Default print statement for mnps class

Description

Default print statement for mnps class

Usage

## S3 method for class 'mnps'
print(x, ...)

Arguments

Default print statement for ps class

Description

Default print statement for ps class

Usage

## S3 method for class 'ps'
print(x, ...)

Arguments

Produces a summary table for iptw object

Description

Produces a summary table for iptw object

Usage

## S3 method for class 'summary.iptw'
print(x, ...)

Arguments

Produces a summary table for mniptw object

Description

Produces a summary table for mniptw object

Usage

## S3 method for class 'summary.mniptw'
print(x, ...)

Arguments

Produces a summary table for mnps object

Description

Produces a summary table for mnps object

Usage

## S3 method for class 'summary.mnps'
print(x, ...)

Arguments

Produces a summary table for ps object

Description

Produces a summary table for ps object

Usage

## S3 method for class 'summary.ps'
print(x, ...)

Arguments

Gradient boosted propensity score estimation

Description

ps calculates propensity scores using gradient boosted logistic regression and diagnoses the resulting propensity scores using a variety of methods

Usage

ps(
  formula = formula(data),
  data,
  n.trees = 10000,
  interaction.depth = 3,
  shrinkage = 0.01,
  bag.fraction = 1,
  n.minobsinnode = 10,
  perm.test.iters = 0,
  print.level = 2,
  verbose = TRUE,
  estimand = "ATE",
  stop.method = c("ks.mean", "es.mean"),
  sampw = NULL,
  version = "gbm",
  ks.exact = NULL,
  n.keep = 1,
  n.grid = 25,
  keep.data = TRUE,
  ...
)

Arguments

Details

For user more comfortable with the options of xgboost::xgboost(), the options for ps controlling the behavior of the gradient boosting algorithm can be specified using the xgboost naming scheme. This includes nrounds, max_depth, eta, and subsample. In addition, the list of parameters passed to xgboost can be specified with params.

Note that unlike earlier versions of 'twang', the plotting functions are no longer included in the ps function. See plot for details of the plots.

Value

Returns an object of class ps, a list containing

gbm.obj

The returned gbm or xgboost object.

treat

The vector of treatment indicators.

treat.var

The treatment variable.

desc

A list containing balance tables for each method selected in stop.methods. Includes a component for the unweighted analysis names “unw”. Each desc component includes a list with the following components

ess

The effective sample size of the control group.

n.treat

The number of subjects in the treatment group.

n.ctrl

The number of subjects in the control group.

max.es

The largest effect size across the covariates.

mean.es

The mean absolute effect size.

max.ks

The largest KS statistic across the covariates.

mean.ks

The average KS statistic across the covariates.

bal.tab

a (potentially large) table summarizing the quality of the weights for equalizing the distribution of features across the two groups. This table is best extracted using the bal.table method. See the help for bal.table for details on the table's contents.

n.trees

The estimated optimal number of gradient boosted iterations to optimize the loss function for the associated stop.methods.

ps

a data frame containing the estimated propensity scores. Each column is associated with one of the methods selected in stop.methods.

w

a data frame containing the propensity score weights. Each column is associated with one of the methods selected in stop.methods.

If sampling weights are given then these are incorporated into these weights.

estimand

The estimand of interest (ATT or ATE).

datestamp

Records the date of the analysis.

parameters

Saves the ps call.

alerts

Text containing any warnings accumulated during the estimation.

iters

A sequence of iterations used in the GBM fits used by plot function.

balance

The balance measures for the pretreatment covariates used in plotting, with a column for each stop.method.

balance.ks

The KS balance measures for the pretreatment covariates used in plotting, with a column for each covariate.

balance.es

The standard differences for the pretreatment covariates used in plotting, with a column for each covariate.

ks

The KS balance measures for the pretreatment covariates on a finer grid, with a column for each covariate.

es

The standard differences for the pretreatment covariates on a finer grid, with a column for each covariate.

n.trees

Maximum number of trees considered in GBM fit.

data

Data as specified in the data argument.

References

Dan McCaffrey, G. Ridgeway, Andrew Morral (2004). "Propensity Score Estimation with Boosted Regression for Evaluating Adolescent Substance Abuse Treatment", *Psychological Methods* 9(4):403-425.

See Also

gbm, xgboost, plot, bal.table

Traffic stop data

Description

Simulated example data for assessing race bias in traffic stop outcomes

Usage

data(raceprofiling)

Format

A data frame with 5000 observations on the following 10 variables.

id

an ID for each traffic stop

nhood

a factor indicating the neighborhood in which the stop occurred.

reason

The reason for the stop, mechanical/registration violations, dangerous moving violation, non-dangerous moving violation

resident

an indicator whether the driver is a resident of the city

age

driver's age

male

an indicator whether the driver was male

race

the race of the driver, with levels A, B, H, W

hour

the hour of the stop (24-hour clock)

month

and ordered factor indicating in which month the stop took place

citation

an indicator of whether the driver received a citation

Source

This is simulated data to demonstrate how to use twang to adjust estimates of racial bias for important factors. This dataset does not represent real data from any real law enforcement agency.

References

G. Ridgeway (2006). “Assessing the effect of race bias in post-traffic stop outcomes using propensity scores,” Journal of Quantitative Criminology 22(1).

Examples

data(raceprofiling)

# the first five lines of the dataset
raceprofiling[1:5,]

Function to run sensitivity analysis described in Ridgeway's paper; currently works only for ATT.

Description

Performs the sensitivity analyses described in Ridgeway (2006). This is a beta version of this functionality. Please let the developers know if you have problems with it.

Usage

sensitivity(ps1, data, outcome, order.by.importance = TRUE, verbose = TRUE)

Arguments

Value

Returns the following * 'tx' Summary for treated observations. * 'ctrl' Summary for control observations.

References

Ridgeway, G. (2006). "The effect of race bias in post-traffic stop outcomes using propensity scores", *Journal of Quantitative Criminology* 22(1):1-29.

Stop methods (e.g. "es.mean", "ks.mean", etc.) object, used only for backward compatibility

Description

In older versions of twang, the 'ps' function specified the 'stop.method' in a different manner. This 'stop.methods' object is used to ensure backward compatibility; new twang users should not make use of it.

Usage

stop.methods

Format

An object of class matrix (inherits from array) with 1 rows and 6 columns.

Details

This is merely a vector with the names of the stopping rules.

Summarize a iptw object

Description

Computes summary information about a stored iptw object

Usage

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

Arguments

Details

Compresses the information in the desc component of the iptw object into a short summary table describing the size of the dataset and the quality of the propensity score weights.

Value

See iptw for details on the returned table.

See Also

iptw

Summarize a mniptw object

Description

Summarize a mniptw object

Usage

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

Arguments

Summarize a mnps object

Description

Computes summary information about a stored mnps object

Usage

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

Arguments

Details

Compresses the information in the desc component of the mnps object into a short summary table describing the size of the dataset and the quality of the propensity score weights.

Value

See mnps for details on the returned table.

See Also

mnps

Summarize a ps object

Description

Computes summary information about a stored ps object

Usage

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

Arguments

Details

Compresses the information in the desc component of the ps object into a short summary table describing the size of the dataset and the quality of the propensity score weights.

Value

See ps for details on the returned table.

See Also

ps