Package 'rdmulti'

Title: Robust Local Polynomial Methods for RD Designs with Multiple Cutoffs or Multiple Scores
Description: The 'rdmulti' package implements estimation, inference, and graphical procedures for regression discontinuity (RD) designs with multiple cutoffs or multiple scores. rdmc() provides point estimation and robust bias-corrected inference for multi-cutoff designs, rdmcplot() provides data-driven RD plots for multi-cutoff designs, and rdms() provides point estimation and robust bias-corrected inference for multi-score designs. See Cattaneo, Titiunik and Vazquez-Bare (2020) <https://rdpackages.github.io/references/Cattaneo-Titiunik-VazquezBare_2020_Stata.pdf> for further methodological details.
Authors: Matias D. Cattaneo [aut, cre], Rocio Titiunik [aut], Gonzalo Vazquez-Bare [aut]
Maintainer: Matias D. Cattaneo <[email protected]>
License: GPL-3
Version: 2.0.0
Built: 2026-05-18 11:12:31 UTC
Source: https://github.com/cran/rdmulti

Help Index

Robust Local Polynomial Methods for RD designs with Multiple Cutoffs or Multiple Scores

Description

The package rdmulti implements estimation, inference, and graphical procedures for Regression Discontinuity (RD) designs with multiple cutoffs or multiple scores. rdmc() provides point estimation and robust bias-corrected inference for multi-cutoff designs, rdmcplot() provides data-driven RD plots for multi-cutoff designs, and rdms() provides point estimation and robust bias-corrected inference for multi-score designs. For more details, and related Stata and R packages useful for analysis of RD designs, visit https://rdpackages.github.io/.

Author(s)

Matias D. Cattaneo, Princeton University. [email protected]

Rocio Titiunik, Princeton University. [email protected]

Gonzalo Vazquez-Bare, UC Santa Barbara. [email protected]

References

Calonico, S., M.D. Cattaneo, M. Farrell and R. Titiunik. (2017). rdrobust: Software for Regression Discontinuity Designs. Stata Journal 17(2): 372-404.

Calonico, S., M.D. Cattaneo, and R. Titiunik. (2014). Robust Data-Driven Inference in the Regression-Discontinuity Design. Stata Journal 14(4): 909-946.

Calonico, S., M.D. Cattaneo, and R. Titiunik. (2015). rdrobust: An R Package for Robust Nonparametric Inference in Regression-Discontinuity Designs. R Journal 7(1): 38-51.

Cattaneo, M.D., L. Keele, R. Titiunik and G. Vazquez-Bare. (2016). Interpreting Regression Discontinuity Designs with Multiple Cutoffs. Journal of Politics 78(4): 1229-1248.

Cattaneo, M.D., L. Keele, R. Titiunik and G. Vazquez-Bare. (2020). Extrapolating Treatment Effects in Multi-Cutoff Regression Discontinuity Designs. Journal of the American Statistical Association 116(536): 1941, 1952.

Cattaneo, M.D., R. Titiunik and G. Vazquez-Bare. (2020). Analysis of Regression Discontinuity Designs with Multiple Cutoffs or Multiple Scores. Stata Journal 20(4): 866-891.

Keele, L. and R. Titiunik. (2015). Geographic Boundaries as Regression Discontinuities. Political Analysis 23(1): 127-155

See Also

Useful links:

Point estimation and robust bias-corrected inference for multi-cutoff designs

Description

rdmc() implements point estimation and robust bias-corrected inference for Regression Discontinuity (RD) designs with multiple cutoffs.

Usage

rdmc(
  Y,
  X,
  C,
  fuzzy = NULL,
  derivvec = NULL,
  pooled_opt = NULL,
  verbose = FALSE,
  pvec = NULL,
  qvec = NULL,
  hmat = NULL,
  bmat = NULL,
  rhovec = NULL,
  covs_mat = NULL,
  covs_list = NULL,
  covs_dropvec = NULL,
  kernelvec = NULL,
  weightsvec = NULL,
  bwselectvec = NULL,
  scaleparvec = NULL,
  scaleregulvec = NULL,
  masspointsvec = NULL,
  bwcheckvec = NULL,
  bwrestrictvec = NULL,
  stdvarsvec = NULL,
  vcevec = NULL,
  nnmatchvec = NULL,
  cluster = NULL,
  ginv.tolvec = NULL,
  sharpbwvec = NULL,
  level = 95,
  plot = FALSE,
  conventional = FALSE,
  subset = NULL,
  data = NULL
)

Arguments

Y

outcome variable.
X

running variable.
C

cutoff variable.
fuzzy

specifies a fuzzy design. See rdrobust() for details.
derivvec

vector of cutoff-specific order of derivatives. See rdrobust() for details.
pooled_opt

options to be passed to rdrobust() to calculate pooled estimand.
verbose

displays the output from rdrobust for estimating the pooled estimand.
pvec

vector of cutoff-specific polynomial orders. See rdrobust() for details.
qvec

vector of cutoff-specific polynomial orders for bias estimation. See rdrobust() for details.
hmat

matrix of cutoff-specific bandwidths. See rdrobust() for details.
bmat

matrix of cutoff-specific bandwidths for bias estimation. See rdrobust() for details.
rhovec

vector of cutoff-specific values of rho. See rdrobust() for details.
covs_mat

matrix of covariates. See rdrobust() for details.
covs_list

list of covariates to be used in each cutoff.
covs_dropvec

vector indicating whether collinear covariates should be dropped at each cutoff. See rdrobust() for details.
kernelvec

vector of cutoff-specific kernels. See rdrobust() for details.
weightsvec

vector of length equal to the number of cutoffs indicating the names of the variables to be used as weights in each cutoff. See rdrobust() for details.
bwselectvec

vector of cutoff-specific bandwidth selection methods. See rdrobust() for details.
scaleparvec

vector of cutoff-specific scale parameters. See rdrobust() for details.
scaleregulvec

vector of cutoff-specific scale regularization parameters. See rdrobust() for details.
masspointsvec

vector indicating how to handle repeated values at each cutoff. See rdrobust() for details.
bwcheckvec

vector indicating the value of bwcheck at each cutoff. See rdrobust() for details.
bwrestrictvec

vector indicating whether computed bandwidths are restricted to the range or runvar at each cutoff. See rdrobust() for details.
stdvarsvec

vector indicating whether variables are standardized at each cutoff. See rdrobust() for details.
vcevec

vector of cutoff-specific variance-covariance estimation methods. See rdrobust() for details.
nnmatchvec

vector of cutoff-specific nearest neighbors for variance estimation. See rdrobust() for details.
cluster

cluster ID variable. See rdrobust() for details.
ginv.tolvec

vector of cutoff-specific tolerances for generalized inverse calculations. See rdrobust() for details.
sharpbwvec

vector indicating whether fuzzy RD bandwidth selection uses the sharp RD model at each cutoff. See rdrobust() for details.
level

confidence level for confidence intervals. See rdrobust() for details.
plot

plots cutoff-specific estimates and weights.
conventional

reports conventional, instead of robust-bias corrected, p-values and confidence intervals.
subset

optional subset of observations to use.
data

optional data frame for resolving string variable names.

Value

tau

pooled estimate
se.rb

robust bias corrected standard error for pooled estimate
pv.rb

robust bias corrected p-value for pooled estimate
ci.rb.l

left limit of robust bias corrected CI for pooled estimate
ci.rb.r

right limit of robust bias corrected CI for pooled estimate
hl

bandwidth to the left of the cutoff for pooled estimate
hr

bandwidth to the right of the cutofffor pooled estimate
Nhl

sample size within bandwidth to the left of the cutoff for pooled estimate
Nhr

sample size within bandwidth to the right of the cutoff for pooled estimate
B

vector of bias-corrected estimates
V

vector of robust variances of the estimates
Coefs

vector of conventional estimates
W

vector of weights for each cutoff-specific estimate
Nh

vector of sample sizes within bandwidth
CI

robust bias-corrected confidence intervals
H

matrix of bandwidths
Pv

vector of robust p-values
rdrobust.results

results from rdrobust for pooled estimate
cfail

Cutoffs where rdrobust() encountered problems

Author(s)

Matias D. Cattaneo, Princeton University. [email protected]

Rocio Titiunik, Princeton University. [email protected]

Gonzalo Vazquez-Bare, UC Santa Barbara. [email protected]

References

Cattaneo, M.D., R. Titiunik and G. Vazquez-Bare. (2020). Analysis of Regression Discontinuity Designs with Multiple Cutoffs or Multiple Scores. Stata Journal, forthcoming.

Examples

# Toy dataset
X <- runif(1000,0,100)
C <- c(rep(33,500),rep(66,500))
Y <- (1 + X + (X>=C))*(C==33)+(.5 + .5*X + .8*(X>=C))*(C==66) + rnorm(1000)
# rdmc with standard syntax
tmp <- rdmc(Y,X,C)

Data-driven RD plots for multi-cutoff designs

Description

rdmcplot() implements data-driven Regression Discontinuity (RD) plots for designs with multiple cutoffs.

Usage

rdmcplot(
  Y,
  X,
  C,
  nbinsmat = NULL,
  binselectvec = NULL,
  scalevec = NULL,
  supportmat = NULL,
  pvec = NULL,
  hmat = NULL,
  kernelvec = NULL,
  weightsvec = NULL,
  covs_mat = NULL,
  covs_list = NULL,
  covs_evalvec = NULL,
  covs_dropvec = NULL,
  ginv.tolvec = NULL,
  masspointsvec = NULL,
  ci = NULL,
  shade = FALSE,
  col_bins = NULL,
  pch_bins = NULL,
  col_poly = NULL,
  lty_poly = NULL,
  col_xline = NULL,
  lty_xline = NULL,
  nobins = FALSE,
  nopoly = FALSE,
  noxline = FALSE,
  nodraw = FALSE,
  subset = NULL,
  data = NULL
)

Arguments

Y

outcome variable.
X

running variable.
C

cutoff variable.
nbinsmat

matrix of cutoff-specific number of bins. See rdplot() for details.
binselectvec

vector of cutoff-specific bins selection method. See rdplot() for details.
scalevec

vector of cutoff-specific scale factors. See rdplot() for details.
supportmat

matrix of cutoff-specific support conditions. See rdplot() for details..
pvec

vector of cutoff-specific polynomial orders. See rdplot() for details.
hmat

matrix of cutoff-specific bandwidths. See rdplot() for details.
kernelvec

vector of cutoff-specific kernels. See rdplot() for details.
weightsvec

vector of cutoff-specific weights. See rdplot() for details.
covs_mat

matrix of covariates. See rdplot() for details.
covs_list

list of of covariates to be used in each cutoff.
covs_evalvec

vector indicating the evaluation point for additional covariates. See rdrobust() for details.
covs_dropvec

vector indicating whether collinear covariates should be dropped at each cutoff. See rdrobust() for details.
ginv.tolvec

vector of cutoff-specific tolerances for generalized inverse calculations. See rdplot() for details.
masspointsvec

vector indicating how to handle repeated values at each cutoff. See rdplot() for details.
ci

adds confidence intervals of the specified level to the plot. See rdrobust() for details.
shade

passes the shaded confidence interval option to rdplot().
col_bins

vector of colors for bins.
pch_bins

vector of characters (pch) type for bins.
col_poly

vector of colors for polynomial curves.
lty_poly

vector of lty for polynomial curves.
col_xline

vector of colors for vertical lines.
lty_xline

vector of lty for vertical lines.
nobins

omits bins plot.
nopoly

omits polynomial curve plot.
noxline

omits vertical lines indicating the cutoffs.
nodraw

omits plot.
subset

optional subset of observations to use.
data

optional data frame for resolving string variable names.

Value

clist

list of cutoffs
cnum

number of cutoffs
X0

matrix of X values for control units
X1

matrix of X values for treated units
Yhat0

estimated polynomial for control units
Yhat1

estimated polynomial for treated units
Xmean

bin average of X values
Ymean

bin average for Y values
CI_l

lower end of confidence intervals
CI_r

upper end of confidence intervals
cfail

Cutoffs where rdrobust() encountered problems

Author(s)

Matias D. Cattaneo, Princeton University. [email protected]

Rocio Titiunik, Princeton University. [email protected]

Gonzalo Vazquez-Bare, UC Santa Barbara. [email protected]

References

Cattaneo, M.D., R. Titiunik and G. Vazquez-Bare. (2020). Analysis of Regression Discontinuity Designs with Multiple Cutoffs or Multiple Scores. Stata Journal, forthcoming.

Examples

# Toy dataset
X <- runif(1000,0,100)
C <- c(rep(33,500),rep(66,500))
Y <- (1 + X + (X>=C))*(C==33)+(.5 + .5*X + .8*(X>=C))*(C==66) + rnorm(1000)
# rdmcplot with standard syntax
tmp <- rdmcplot(Y,X,C)

Point estimation and robust bias-corrected inference for multi-score designs

Description

rdms() implements point estimation and robust bias-corrected inference for Regression Discontinuity (RD) designs with multiple scores, including cumulative-cutoff designs and designs with two running variables.

Usage

rdms(
  Y,
  X,
  C,
  X2 = NULL,
  zvar = NULL,
  C2 = NULL,
  rangemat = NULL,
  xnorm = NULL,
  fuzzy = NULL,
  derivvec = NULL,
  pooled_opt = NULL,
  pvec = NULL,
  qvec = NULL,
  hmat = NULL,
  bmat = NULL,
  rhovec = NULL,
  covs_mat = NULL,
  covs_list = NULL,
  covs_dropvec = NULL,
  kernelvec = NULL,
  weightsvec = NULL,
  bwselectvec = NULL,
  scaleparvec = NULL,
  scaleregulvec = NULL,
  masspointsvec = NULL,
  bwcheckvec = NULL,
  bwrestrictvec = NULL,
  stdvarsvec = NULL,
  vcevec = NULL,
  nnmatchvec = NULL,
  cluster = NULL,
  ginv.tolvec = NULL,
  sharpbwvec = NULL,
  level = 95,
  plot = FALSE,
  conventional = FALSE,
  subset = NULL,
  data = NULL
)

Arguments

Y

outcome variable.
X

running variable.
C

vector of cutoffs.
X2

if specified, second running variable.
zvar

if X2 is specified, treatment indicator.
C2

if specified, second vector of cutoffs.
rangemat

matrix of cutoff-specific ranges for the running variable.
xnorm

normalized running variable to estimate pooled effect.
fuzzy

specifies a fuzzy design. See rdrobust() for details.
derivvec

vector of cutoff-specific order of derivatives. See rdrobust() for details.
pooled_opt

options to be passed to rdrobust() to calculate pooled estimand.
pvec

vector of cutoff-specific polynomial orders. See rdrobust() for details.
qvec

vector of cutoff-specific polynomial orders for bias estimation. See rdrobust() for details.
hmat

matrix of cutoff-specific bandwidths. See rdrobust() for details.
bmat

matrix of cutoff-specific bandwidths for bias estimation. See rdrobust() for details.
rhovec

vector of cutoff-specific values of rho. See rdrobust() for details.
covs_mat

matrix of covariates. See rdplot() for details.
covs_list

list of of covariates to be used in each cutoff.
covs_dropvec

vector indicating whether collinear covariates should be dropped at each cutoff. See rdrobust() for details.
kernelvec

vector of cutoff-specific kernels. See rdrobust() for details.
weightsvec

vector of length equal to the number of cutoffs indicating the names of the variables to be used as weights in each cutoff. See rdrobust() for details.
bwselectvec

vector of cutoff-specific bandwidth selection methods. See rdrobust() for details.
scaleparvec

vector of cutoff-specific scale parameters. See rdrobust() for details.
scaleregulvec

vector of cutoff-specific scale regularization parameters. See rdrobust() for details.
masspointsvec

vector indicating how to handle repeated values at each cutoff. See rdrobust() for details.
bwcheckvec

vector indicating the value of bwcheck at each cutoff. See rdrobust() for details.
bwrestrictvec

vector indicating whether computed bandwidths are restricted to the range or runvar at each cutoff. See rdrobust() for details.
stdvarsvec

vector indicating whether variables are standardized at each cutoff. See rdrobust() for details.
vcevec

vector of cutoff-specific variance-covariance estimation methods. See rdrobust() for details.
nnmatchvec

vector of cutoff-specific nearest neighbors for variance estimation. See rdrobust() for details.
cluster

cluster ID variable. See rdrobust() for details.
ginv.tolvec

vector of cutoff-specific tolerances for generalized inverse calculations. See rdrobust() for details.
sharpbwvec

vector indicating whether fuzzy RD bandwidth selection uses the sharp RD model at each cutoff. See rdrobust() for details.
level

confidence level for confidence intervals. See rdrobust() for details.
plot

plots cutoff-specific and pooled estimates.
conventional

reports conventional, instead of robust-bias corrected, p-values and confidence intervals.
subset

optional subset of observations to use.
data

optional data frame for resolving string variable names.

Value

B

vector of bias-corrected coefficients
V

variance-covariance matrix of the estimators
Coefs

vector of conventional coefficients
Nh

vector of sample sizes within bandwidth at each cutoff
CI

bias corrected confidence intervals
H

bandwidth used at each cutoff
Pv

vector of robust p-values

Author(s)

Matias D. Cattaneo, Princeton University. [email protected]

Rocio Titiunik, Princeton University. [email protected]

Gonzalo Vazquez-Bare, UC Santa Barbara. [email protected]

References

Cattaneo, M.D., R. Titiunik and G. Vazquez-Bare. (2020). Analysis of Regression Discontinuity Designs with Multiple Cutoffs or Multiple Scores. Stata Journal, forthcoming.

Examples

# Toy dataset: cumulative cutoffs
X <- runif(1000,0,100)
C <- c(33,66)
Y <- (1+X)*(X<C[1])+(0.8+0.8*X)*(X>=C[1]&X<C[2])+(1.2+1.2*X)*(X>=C[2]) + rnorm(1000)
# rmds: basic syntax
tmp <- rdms(Y,X,C)