Skip to content

Commit

Permalink
Pull Request follow-up for #30: Various minor fixes (#42)
Browse files Browse the repository at this point in the history 
* Added 'value' to man/constructOriginalCS.Rd

* Closed [ in description

* Changed all T/F to TRUE/FALSE

* Removed all standardtext from .Rd files

* Give Ashesh and Jon author and copyright; Ashesh creator/maintainer

* Added simple examples for main user funs

* Minor fix

* Reference vignette name directly in simple examples

* Direct users to github README in help files
  • Loading branch information
@mcaceresb
mcaceresb authored Oct 22, 2023
1 parent bb2c7b6 commit 10c13d6
Show file tree
Hide file tree
Showing 35 changed files with 213 additions and 159 deletions.
13 changes: 10 additions & 3 deletions DESCRIPTION
View file
Original file line number Diff line number Diff line change
Expand Up @@ -24,10 +24,17 @@ Imports:
Suggests:
knitr,
rmarkdown
Author: Ashesh Rambachan <[email protected]>
Maintainer: Ashesh Rambachan <[email protected]>
Authors@R:
c(person(given = "Ashesh",
family = "Rambachan",
role = c("aut", "cph", "cre"),
email = "[email protected]"),
person(given = "Jonathan",
family = "Roth",
role = c("aut", "cph"),
email = "[email protected]"))
Description:
Provides functions to conduct robust inference in difference-in-differences and event study designs by implementing the methods developed in Rambachan & Roth (2023, RESTUD), "A More Credible Approach to Parallel Trends" [Previously titled "An Honest Approach...". Inference is conducted under a weaker version of the parallel trends assumption. Uniformly valid confidence sets are constructed based upon conditional confidence sets, fixed-length confidence sets and hybridized confidence sets.
Provides functions to conduct robust inference in difference-in-differences and event study designs by implementing the methods developed in Rambachan & Roth (2023, RESTUD), "A More Credible Approach to Parallel Trends" [Previously titled "An Honest Approach..."]. Inference is conducted under a weaker version of the parallel trends assumption. Uniformly valid confidence sets are constructed based upon conditional confidence sets, fixed-length confidence sets and hybridized confidence sets.
Encoding: UTF-8
LazyData: true
VignetteBuilder:
Expand Down
2 changes: 1 addition & 1 deletion R/HonestDiD-Temp.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -416,7 +416,7 @@ createEventStudyPlot <- function(betahat, stdErrors = NULL, sigma = NULL,
stdErrors = sqrt(diag(sigma))
}

if (useRelativeEventTime == T) {
if (useRelativeEventTime == TRUE) {
timeVec = timeVec - referencePeriod
referencePeriod = 0
}
Expand Down
4 changes: 2 additions & 2 deletions R/arp-nonuisance.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -70,7 +70,7 @@

# Run the first-stage test
if (base::max(A_firststage %*% y - d_firststage) > 0) {
reject <- T
reject <- TRUE
} else {
# Per ARP (2021), CV = max(0, c_{1-alpha-tilde}), where alpha-tilde = (alpha - kappa)/(1-kappa)
# quantile of truncated normal that accounts for failing to reject in the first stage.
Expand Down Expand Up @@ -181,7 +181,7 @@
}

# Compute length, else return grid
if (returnLength == T) {
if (returnLength == TRUE) {
gridLength <- 0.5 * ( base::c(0, base::diff(thetaGrid)) + base::c(base::diff(thetaGrid), 0 ) )
base::return(base::sum(resultsGrid[, 2]*gridLength))
} else {
Expand Down
8 changes: 4 additions & 4 deletions R/arp-nuisance.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -400,7 +400,7 @@
eta_vec = base::apply(-xi.draws, 1, .compute_eta, f, C)

# We compute the 1-kappa quantile of eta_vec and return this value
base::return(stats::quantile(eta_vec, probs=1-hybrid_kappa, names=FALSE, na.rm=T))
base::return(stats::quantile(eta_vec, probs=1-hybrid_kappa, names=FALSE, na.rm=TRUE))
}
}

Expand Down Expand Up @@ -513,11 +513,11 @@

# Store which moments are binding: Do so using lambda rather than the moments.
B_index = (linSoln$lambda > tol_lambda)
Bc_index = (B_index == F)
Bc_index = (B_index == FALSE)
X_TB = base::matrix( X_T_ARP[B_index,], ncol = base::ncol(X_T_ARP) ) # select binding moments
# Check whether binding moments have full rank.
if (base::is.vector(X_TB)) {
fullRank_flag = F
fullRank_flag = FALSE
} else {
fullRank_flag = (Matrix::rankMatrix(X_TB) == base::min(base::dim( X_TB ) ))
}
Expand Down Expand Up @@ -761,7 +761,7 @@
}

# Compute length, else return grid
if (returnLength == T) {
if (returnLength == TRUE) {
gridLength <- 0.5 * ( base::c(0, base::diff(thetaGrid)) + base::c(base::diff(thetaGrid), 0 ) )
base::return(base::sum(resultsGrid[, 2]*gridLength))
} else {
Expand Down
2 changes: 1 addition & 1 deletion R/delta_utility_functions.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@
}
}

# If postPeriodMomentsOnly == T, exclude moments that only involve pre-periods
# If postPeriodMomentsOnly == TRUE, exclude moments that only involve pre-periods
if(postPeriodMomentsOnly){
postPeriodIndices <- (numPrePeriods +1):base::NCOL(A_M)
prePeriodOnlyRows <- base::which( base::rowSums( A_M[ , postPeriodIndices] != 0 ) == 0 )
Expand Down
18 changes: 9 additions & 9 deletions R/deltarm.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -10,10 +10,10 @@

# Delta^{RM} functions -----------------------------------------------
.create_A_RM <- function(numPrePeriods, numPostPeriods,
Mbar = 1, s, max_positive = T,
dropZero = T) {
Mbar = 1, s, max_positive = TRUE,
dropZero = TRUE) {
# This function creates a matrix for the linear constraints that
# \delta \in Delta^RM_{s,(.)}(Mbar), where (.) is + if max_positve = T and (-) if max_positive = F.
# \delta \in Delta^RM_{s,(.)}(Mbar), where (.) is + if max_positve = TRUE and (-) if max_positive = FALSE.
#
# Inputs:
# numPrePeriods = number of pre-periods. This is an element of resultsObjects.
Expand All @@ -27,11 +27,11 @@
Atilde[r, r:(r+1)] = c(-1, 1)
}

# Create a vector to extract the max first dif, which corresponds with the first dif for period s, or minus this if max_positive == F
# Create a vector to extract the max first dif, which corresponds with the first dif for period s, or minus this if max_positive == FALSE
v_max_dif <- base::matrix(0, nrow = 1, ncol = numPrePeriods + numPostPeriods + 1)
v_max_dif[(numPrePeriods+s):(numPrePeriods+1+s)] <- c(-1,1)

if (max_positive == F){
if (max_positive == FALSE){
v_max_dif <- -v_max_dif
}

Expand All @@ -55,9 +55,9 @@
}
}

.create_d_RM <- function(numPrePeriods, numPostPeriods, dropZero = T){
.create_d_RM <- function(numPrePeriods, numPostPeriods, dropZero = TRUE){
# This function creates a vector for the linear constraints that
# delta is in Delta^RM_{s,(.)}(Mbar), where (.) is + if max_positve = T and - if max_positive = F.
# delta is in Delta^RM_{s,(.)}(Mbar), where (.) is + if max_positve = TRUE and - if max_positive = FALSE.
# It implements this using the general characterization of d, NOT the sharp
# characterization of the identified set.
#
Expand Down Expand Up @@ -241,7 +241,7 @@
computeConditionalCS_DeltaRM <- function(betahat, sigma, numPrePeriods, numPostPeriods,
l_vec = .basisVector(index = 1, size = numPostPeriods), Mbar = 0,
alpha = 0.05, hybrid_flag = "LF", hybrid_kappa = alpha/10,
returnLength = F, postPeriodMomentsOnly = T,
returnLength = FALSE, postPeriodMomentsOnly = TRUE,
gridPoints = 10^3, grid.ub = NA, grid.lb = NA) {
# This function computes the ARP CI that includes nuisance parameters
# for Delta^{RM}(Mbar). This functions uses ARP_computeCI for all
Expand Down Expand Up @@ -304,7 +304,7 @@ computeConditionalCS_DeltaRM <- function(betahat, sigma, numPrePeriods, numPostP
CI_RM = tibble::tibble(grid = base::seq(grid.lb, grid.ub, length.out = gridPoints),
accept = base::pmax(CIs_RM_plus_maxS, CIs_RM_minus_maxS))

# Compute length if returnLength == T, else return grid
# Compute length if returnLength == TRUE, else return grid
if (returnLength) {
gridLength <- 0.5 * ( base::c(0, base::diff(CI_RM$grid)) + base::c(base::diff(CI_RM$grid), 0 ) )
base::return(base::sum(CI_RM$accept*gridLength))
Expand Down
20 changes: 10 additions & 10 deletions R/deltarmb.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -12,10 +12,10 @@

# Delta^{RMB} functions -----------------------------------------------
.create_A_RMB <- function(numPrePeriods, numPostPeriods,
Mbar = 1, s, max_positive = T,
dropZero = T, biasDirection) {
Mbar = 1, s, max_positive = TRUE,
dropZero = TRUE, biasDirection) {
# This function creates a matrix for the linear constraints that
# \delta \in Delta^RMB_{s,(.)}(Mbar), where (.) is + if max_positve = T and (-) if max_positive = F.
# \delta \in Delta^RMB_{s,(.)}(Mbar), where (.) is + if max_positve = TRUE and (-) if max_positive = FALSE.
#
# Inputs:
# numPrePeriods = number of pre-periods. This is an element of resultsObjects.
Expand All @@ -30,11 +30,11 @@
Atilde[r, r:(r+1)] = base::c(-1, 1)
}

# Create a vector to extract the max first dif, which corresponds with the first dif for period s, or minus this if max_positive == F
# Create a vector to extract the max first dif, which corresponds with the first dif for period s, or minus this if max_positive == FALSE
v_max_dif <- base::matrix(0, nrow = 1, ncol = numPrePeriods + numPostPeriods + 1)
v_max_dif[(numPrePeriods+s):(numPrePeriods+1+s)] <- c(-1,1)

if (max_positive == F){
if (max_positive == FALSE){
v_max_dif <- -v_max_dif
}

Expand Down Expand Up @@ -65,9 +65,9 @@
}
}

.create_d_RMB <- function(numPrePeriods, numPostPeriods, dropZero = T){
.create_d_RMB <- function(numPrePeriods, numPostPeriods, dropZero = TRUE){
# This function creates a vector for the linear constraints that
# delta is in Delta^RMB_{s,(.)}(Mbar), where (.) is + if max_positve = T and - if max_positive = F.
# delta is in Delta^RMB_{s,(.)}(Mbar), where (.) is + if max_positve = TRUE and - if max_positive = FALSE.
# It implements this using the general characterization of d, NOT the sharp
# characterization of the identified set.
#
Expand Down Expand Up @@ -258,8 +258,8 @@
computeConditionalCS_DeltaRMB <- function(betahat, sigma, numPrePeriods, numPostPeriods,
l_vec = .basisVector(index = 1, size = numPostPeriods), Mbar = 0,
alpha = 0.05, hybrid_flag = "LF", hybrid_kappa = alpha/10,
returnLength = F, biasDirection = "positive",
postPeriodMomentsOnly = T,
returnLength = FALSE, biasDirection = "positive",
postPeriodMomentsOnly = TRUE,
gridPoints = 10^3, grid.ub = NA, grid.lb = NA) {
# This function computes the ARP CI that includes nuisance parameters
# for Delta^{MB}(Mbar). This functions uses ARP_computeCI for all
Expand Down Expand Up @@ -323,7 +323,7 @@ computeConditionalCS_DeltaRMB <- function(betahat, sigma, numPrePeriods, numPost
CI_RMB = tibble::tibble(grid = base::seq(grid.lb, grid.ub, length.out = gridPoints),
accept = base::pmax(CIs_RMB_plus_maxS, CIs_RMB_minus_maxS))

# Compute length if returnLength == T, else return grid
# Compute length if returnLength == TRUE, else return grid
if (returnLength) {
gridLength <- 0.5 * ( base::c(0, base::diff(CI_RMB$grid)) + base::c(base::diff(CI_RMB$grid), 0 ) )
base::return(base::sum(CI_RMB$accept*gridLength))
Expand Down
18 changes: 9 additions & 9 deletions R/deltarmm.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -11,10 +11,10 @@

# Delta^{RMM} functions -----------------------------------------------
.create_A_RMM <- function(numPrePeriods, numPostPeriods,
Mbar = 1, s, max_positive = T,
dropZero = T, monotonicityDirection) {
Mbar = 1, s, max_positive = TRUE,
dropZero = TRUE, monotonicityDirection) {
# This function creates a matrix for the linear constraints that
# \delta \in Delta^RM_{s,(.)}(Mbar), where (.) is + if max_positve = T and (-) if max_positive = F.
# \delta \in Delta^RM_{s,(.)}(Mbar), where (.) is + if max_positve = TRUE and (-) if max_positive = FALSE.
#
# Inputs:
# numPrePeriods = number of pre-periods. This is an element of resultsObjects.
Expand All @@ -29,11 +29,11 @@
Atilde[r, r:(r+1)] = c(-1, 1)
}

# Create a vector to extract the max first dif, which corresponds with the first dif for period s, or minus this if max_positive == F
# Create a vector to extract the max first dif, which corresponds with the first dif for period s, or minus this if max_positive == FALSE
v_max_dif <- base::matrix(0, nrow = 1, ncol = numPrePeriods + numPostPeriods + 1)
v_max_dif[(numPrePeriods+s):(numPrePeriods+1+s)] <- c(-1,1)

if (max_positive == F){
if (max_positive == FALSE){
v_max_dif <- -v_max_dif
}

Expand Down Expand Up @@ -63,9 +63,9 @@
}
}

.create_d_RMM <- function(numPrePeriods, numPostPeriods, dropZero = T){
.create_d_RMM <- function(numPrePeriods, numPostPeriods, dropZero = TRUE){
# This function creates a vector for the linear constraints that
# delta is in Delta^RM_{s,(.)}(Mbar), where (.) is + if max_positve = T and - if max_positive = F.
# delta is in Delta^RM_{s,(.)}(Mbar), where (.) is + if max_positve = TRUE and - if max_positive = FALSE.
# It implements this using the general characterization of d, NOT the sharp
# characterization of the identified set.
#
Expand Down Expand Up @@ -257,7 +257,7 @@
computeConditionalCS_DeltaRMM <- function(betahat, sigma, numPrePeriods, numPostPeriods,
l_vec = .basisVector(index = 1, size = numPostPeriods), Mbar = 0,
alpha = 0.05, hybrid_flag = "LF", hybrid_kappa = alpha/10,
returnLength = F, postPeriodMomentsOnly = T, monotonicityDirection = "increasing",
returnLength = FALSE, postPeriodMomentsOnly = TRUE, monotonicityDirection = "increasing",
gridPoints = 10^3, grid.ub = NA, grid.lb = NA) {
# This function computes the ARP CI that includes nuisance parameters
# for Delta^{RMM}(Mbar). This functions uses ARP_computeCI for all
Expand Down Expand Up @@ -320,7 +320,7 @@ computeConditionalCS_DeltaRMM <- function(betahat, sigma, numPrePeriods, numPost
CI_RMM = tibble::tibble(grid = base::seq(grid.lb, grid.ub, length.out = gridPoints),
accept = base::pmax(CIs_RMM_plus_maxS, CIs_RMM_minus_maxS))

# Compute length if returnLength == T, else return grid
# Compute length if returnLength == TRUE, else return grid
if (returnLength) {
gridLength <- 0.5 * ( base::c(0, base::diff(CI_RMM$grid)) + base::c(base::diff(CI_RMM$grid), 0 ) )
base::return(base::sum(CI_RMM$accept*gridLength))
Expand Down
14 changes: 7 additions & 7 deletions R/deltasd.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,7 @@
# In this section, we implement helper functions to place testing with
# Delta^{SD}(M) into the form needed to use the ARP functions.

.create_A_SD <- function(numPrePeriods, numPostPeriods, postPeriodMomentsOnly = F) {
.create_A_SD <- function(numPrePeriods, numPostPeriods, postPeriodMomentsOnly = FALSE) {
# This function creates a matrix for the linear constraints that \delta \in Delta^SD(M).
# It implements this using the general characterization of A, NOT the sharp
# characterization of the identified set.
Expand All @@ -31,7 +31,7 @@
}
Atilde = Atilde[, -(numPrePeriods+1)]

# If postPeriodMomentsOnly == T, exclude moments that only involve pre-periods
# If postPeriodMomentsOnly == TRUE, exclude moments that only involve pre-periods
if(postPeriodMomentsOnly){
postPeriodIndices <- (numPrePeriods +1):base::NCOL(Atilde)
prePeriodOnlyRows <- base::which( base::rowSums( Atilde[ , postPeriodIndices] != 0 ) == 0 )
Expand All @@ -43,7 +43,7 @@
base::return(A)
}

.create_d_SD <- function(numPrePeriods, numPostPeriods, M, postPeriodMomentsOnly = F) {
.create_d_SD <- function(numPrePeriods, numPostPeriods, M, postPeriodMomentsOnly = FALSE) {
# This function creates a vector for the linear constraints that \delta \in Delta^SD(M).
# It implements this using the general characterization of d, NOT the sharp
# characterization of the identified set.
Expand Down Expand Up @@ -129,8 +129,8 @@
computeConditionalCS_DeltaSD <- function(betahat, sigma, numPrePeriods, numPostPeriods,
l_vec = .basisVector(index = 1, size = numPostPeriods),
M = 0, alpha = 0.05, hybrid_flag = "FLCI",
hybrid_kappa = alpha/10, returnLength = F,
postPeriodMomentsOnly = T,
hybrid_kappa = alpha/10, returnLength = FALSE,
postPeriodMomentsOnly = TRUE,
gridPoints =10^3, grid.ub = NA, grid.lb = NA) {
# This function computes the ARP CI that includes nuisance parameters
# for Delta^{SD}(M). This functions uses ARP_computeCI for all
Expand All @@ -154,8 +154,8 @@ computeConditionalCS_DeltaSD <- function(betahat, sigma, numPrePeriods, numPostP
# data_frame containing upper and lower bounds of CI.

# Construct A_SD, d_SD
A_SD = .create_A_SD(numPrePeriods = numPrePeriods, numPostPeriods = numPostPeriods, postPeriodMomentsOnly = F)
d_SD = .create_d_SD(numPrePeriods = numPrePeriods, numPostPeriods = numPostPeriods, M = M, postPeriodMomentsOnly = F)
A_SD = .create_A_SD(numPrePeriods = numPrePeriods, numPostPeriods = numPostPeriods, postPeriodMomentsOnly = FALSE)
d_SD = .create_d_SD(numPrePeriods = numPrePeriods, numPostPeriods = numPostPeriods, M = M, postPeriodMomentsOnly = FALSE)

if (postPeriodMomentsOnly & numPostPeriods > 1){
postPeriodIndices <- (numPrePeriods +1):base::NCOL(A_SD)
Expand Down
12 changes: 6 additions & 6 deletions R/deltasdb.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,7 @@
# In this section, we implement helper functions to place testing with
# Delta^{SDB}(M) into the form needed to use the ARP functions.
.create_A_SDB <- function(numPrePeriods, numPostPeriods,
biasDirection = "positive", postPeriodMomentsOnly = F) {
biasDirection = "positive", postPeriodMomentsOnly = FALSE) {
# This function creates a matrix for the linear constraints that \delta \in Delta^SDB(M).
# It implements this using the general characterization of A.
#
Expand All @@ -29,7 +29,7 @@
base::return(A)
}

.create_d_SDB <- function(numPrePeriods, numPostPeriods, M, postPeriodMomentsOnly = F) {
.create_d_SDB <- function(numPrePeriods, numPostPeriods, M, postPeriodMomentsOnly = FALSE) {
# This function creates a vector for the linear constraints that \delta \in Delta^SDB(M).
# It implements this using the general characterization of d.
#
Expand Down Expand Up @@ -112,8 +112,8 @@
computeConditionalCS_DeltaSDB <- function(betahat, sigma, numPrePeriods, numPostPeriods,
M = 0, l_vec = .basisVector(index = 1, size=numPostPeriods),
alpha = 0.05, hybrid_flag = "FLCI", hybrid_kappa = alpha/10,
returnLength = F, biasDirection = "positive",
postPeriodMomentsOnly = T,
returnLength = FALSE, biasDirection = "positive",
postPeriodMomentsOnly = TRUE,
gridPoints = 10^3, grid.lb = NA, grid.ub = NA) {
# This function computes the ARP CI that includes nuisance parameters
# for Delta^{SDPB}(M). This functions uses ARP_computeCI for all
Expand All @@ -140,8 +140,8 @@ computeConditionalCS_DeltaSDB <- function(betahat, sigma, numPrePeriods, numPost
# Construct A_SDB, d_SDB
A_SDB = .create_A_SDB(numPrePeriods = numPrePeriods, numPostPeriods = numPostPeriods,
biasDirection = biasDirection,
postPeriodMomentsOnly = F)
d_SDB = .create_d_SDB(numPrePeriods = numPrePeriods, numPostPeriods = numPostPeriods, M = M, postPeriodMomentsOnly = F)
postPeriodMomentsOnly = FALSE)
d_SDB = .create_d_SDB(numPrePeriods = numPrePeriods, numPostPeriods = numPostPeriods, M = M, postPeriodMomentsOnly = FALSE)

if (postPeriodMomentsOnly & numPostPeriods > 1) {
postPeriodIndices <- (numPrePeriods +1):base::NCOL(A_SDB)
Expand Down
Loading

0 comments on commit 10c13d6

Please sign in to comment.