Package 'senseweight' reference manual

Title:	Sensitivity Analysis for Weighted Estimators
Description:	Provides tools to conduct rigorous sensitivity analyses for weighted estimators, introduced in Huang (2024) <doi:10.1093/jrsssa/qnae012> and Hartman and Huang (2024) <doi:10.1017>. The package allows researchers to generate the set of recommended sensitivity summaries to evaluate the sensitivity in their underlying weighting estimators to omitted moderators or confounders. The tools can be flexibly applied in causal inference settings (i.e., in external and internal validity contexts) or survey contexts.
Authors:	Melody Huang [aut, cre]
Maintainer:	Melody Huang <[email protected]>
License:	MIT + file LICENSE
Version:	0.0.1
Built:	2026-05-20 08:27:58 UTC
Source:	https://github.com/melodyyhuang/senseweight

Benchmark (for survey weights)

Description

Returns benchmarking results for survey weighting

Usage

benchmark_survey(
  omit,
  formula,
  weights,
  pop_svy = NULL,
  sample_svy,
  Y,
  population_targets = NULL,
  weighting_method = "raking"
)
benchmark_survey(
  omit,
  formula,
  weights,
  pop_svy = NULL,
  sample_svy,
  Y,
  population_targets = NULL,
  weighting_method = "raking"
)

Arguments

omit

Variable to benchmark

formula

Raking formula

weights

A vector, containing the estimated survey weights

pop_svy

Survey object, containing the population the survey sample is being re-weighted to

sample_svy

Survey object, containing the survey sample being re-weighted

Y

outcome of interest

population_targets

Population targets for the raking formula (optional, if not provided, will be generated from pop_svy)

weighting_method

Weighting method (default to raking)

Value

Benchmarking results for a variable (or subset of variables)

Examples

data(poll.data)
poll_srs <- survey::svydesign(ids = ~ 1, data = poll.data)
pop_targets = c(1, 0.212, 0.264, 0.236, 0.310, 
                0.114, 0.360, 0.528, 0.114, 
                0.021, 0.034, 0.805, 
                0.266, 0.075, 0.312, 0.349)
names(pop_targets) = c("(Intercept)",
                       "age_buckets36to50",
                       "age_buckets51to64",
                       "age_bucketsOver65",
                       "educHigh School or Less",
                       "educPost-grad",
                       "educSome college",
                       "genderWomen", 
                       "raceBlack",
                       "raceHispanic",
                       "raceOther",
                       "raceWhite", 
                       "pidIndependent", "pidOther", 
                       "pidRepublican", "bornagainYes")
#Set up raking formula:
formula_rake <- ~ age_buckets + educ + gender + race + pid + bornagain

#PERFORM RAKING:
model_rake <- survey::calibrate(
  design = poll_srs,
  formula = formula_rake,
  population = pop_targets,
  calfun = "raking",
  force = TRUE
)


rake_results <- survey::svydesign( ~ 1, data = poll.data, weights = stats::weights(model_rake))
#Estimate from raking results:
weights = stats::weights(rake_results) * nrow(model_rake)

unweighted_estimate = survey::svymean(~ Y, poll_srs, na.rm = TRUE)
weighted_estimate = survey::svymean(~ Y, model_rake, na.rm = TRUE)
benchmark_survey('educ', 
                 formula = formula_rake,
                 weights = weights,
                 population_targets = pop_targets,
                 sample_svy = poll_srs,
                 Y = poll.data$Y)
 
data(poll.data)
poll_srs <- survey::svydesign(ids = ~ 1, data = poll.data)
pop_targets = c(1, 0.212, 0.264, 0.236, 0.310, 
                0.114, 0.360, 0.528, 0.114, 
                0.021, 0.034, 0.805, 
                0.266, 0.075, 0.312, 0.349)
names(pop_targets) = c("(Intercept)",
                       "age_buckets36to50",
                       "age_buckets51to64",
                       "age_bucketsOver65",
                       "educHigh School or Less",
                       "educPost-grad",
                       "educSome college",
                       "genderWomen", 
                       "raceBlack",
                       "raceHispanic",
                       "raceOther",
                       "raceWhite", 
                       "pidIndependent", "pidOther", 
                       "pidRepublican", "bornagainYes")
#Set up raking formula:
formula_rake <- ~ age_buckets + educ + gender + race + pid + bornagain

#PERFORM RAKING:
model_rake <- survey::calibrate(
  design = poll_srs,
  formula = formula_rake,
  population = pop_targets,
  calfun = "raking",
  force = TRUE
)


rake_results <- survey::svydesign( ~ 1, data = poll.data, weights = stats::weights(model_rake))
#Estimate from raking results:
weights = stats::weights(rake_results) * nrow(model_rake)

unweighted_estimate = survey::svymean(~ Y, poll_srs, na.rm = TRUE)
weighted_estimate = survey::svymean(~ Y, model_rake, na.rm = TRUE)
benchmark_survey('educ', 
                 formula = formula_rake,
                 weights = weights,
                 population_targets = pop_targets,
                 sample_svy = poll_srs,
                 Y = poll.data$Y)

Bias Contour Plots

Description

Generates bias contour plots to aid with sensitivity analysis

Usage

contour_plot(
  varW,
  sigma2,
  killer_confounder,
  df_benchmark,
  benchmark = TRUE,
  shade = FALSE,
  shade_var = NULL,
  shade_fill = "#35a4bf",
  shade_alpha = 0.25,
  contour_width = 1,
  binwidth = NULL,
  label_size = 0.25,
  point_size = 1,
  nudge = 0.05,
  axis_text_size = 12,
  axis_title_size = 14,
  axis_line_width = 1,
  print = FALSE
)
contour_plot(
  varW,
  sigma2,
  killer_confounder,
  df_benchmark,
  benchmark = TRUE,
  shade = FALSE,
  shade_var = NULL,
  shade_fill = "#35a4bf",
  shade_alpha = 0.25,
  contour_width = 1,
  binwidth = NULL,
  label_size = 0.25,
  point_size = 1,
  nudge = 0.05,
  axis_text_size = 12,
  axis_title_size = 14,
  axis_line_width = 1,
  print = FALSE
)

Arguments

varW

Variance of the estimated weights

sigma2

Estimated variance of the outcome (i.e., stats::var(Y) for obervational setting; stats::var(tau) for generalization setting)#'

killer_confounder

Threshold for bias considered large enough to be a killer confounder. For example, if researchers are concerned about the bias large enough to reduce an estimated treatment effect to zero or change directional sign, set killer_confounder equal to the point estimate.

df_benchmark

Data frame containing formal benchmarking results. The data.frame must contain the columns variable (for the covariate name), R2_benchmark, and rho_benchmark.

benchmark

Flag for whether or not to display benchmarking results (benchmark = TRUE if we want to add benchmarking results to plot, benchamrk=FALSE otherwise). If set to TRUE, df_benchmark must contain valid benchmarking results.

shade

Flag for whether or not a specific benchmarking covariate (or set of benchmarked covariates) should be shaded a different color (shade = TRUE indicates that we want to highlight specific variables)

shade_var

If shade = TRUE, this contains either a vector containing the variables we want to highlight

shade_fill

Color to fill the highlighted variables. Default is set to "#35a4bf".

shade_alpha

Alpha value for the fill color. Default is set to 0.25.

contour_width

Width of the contour lines. Default is set to 1.

binwidth

If set to a numeric value, the function will generate a contour plot with the specified binwidth. Default is set to NULL.

label_size

Size of the labels. Default is set to 0.25.

point_size

Size of the points. Default is set to 1.

nudge

Nudge value for the labels. Default is set to 0.05.

axis_text_size

Size of the axis text. Default is set to 12.

axis_title_size

Size of the axis title. Default is set to 14.

axis_line_width

Width of the axis lines. Default is set to 1.

print

If set to TRUE, the function will return a list with two elements: plot which contains the generated bias contour plot, and data, which provides the data.frame for generating the contour plot. If set to FALSE, the function will simply generate the bias contour plot. Default is set to FALSE.

Value

A ggplot2 object, containing the bias contour plot

Examples

# For the external validity setting: 
data(jtpa_women)
site_name <- "NE"
df_site <- jtpa_women[which(jtpa_women$site == site_name), ]
df_else <- jtpa_women[which(jtpa_women$site != site_name), ]

# Estimate unweighted estimator:
model_dim <- estimatr::lm_robust(Y ~ T, data = df_site)
PATE <- coef(lm(Y ~ T, data = df_else))[2]
DiM <- coef(model_dim)[2]
# Generate weights using observed covariates:
df_all <- jtpa_women
df_all$S <- ifelse(jtpa_women$site == "NE", 1, 0)
model_ps <- WeightIt::weightit(
  (1 - S) ~ . - site - T - Y, 
  data = df_all, method = "ebal", estimand = "ATT"
)
weights <- model_ps$weights[df_all$S == 1]
# Estimate IPW model:
model_ipw <- estimatr::lm_robust(Y ~ T, data = df_site, weights = weights)
ipw <- coef(model_ipw)[2]
# Estimate bound for var(tau):
vartau <- var(df_site$Y[df_site$T == 1]) - var(df_site$Y[df_site$T == 0])
RV <- robustness_value(estimate = ipw, b_star = 0, sigma2 = vartau, weights = weights)

# Select weighting variables:
weighting_vars <- names(df_all)[which(!names(df_all) %in% c("site", "S", "Y", "T"))]

# Run benchmarking:
df_benchmark <- run_benchmarking(
  weighting_vars = weighting_vars,
  data = df_all[, -1],
  treatment = "T", outcome = "Y", selection = "S",
  estimate = ipw,
  RV = RV, sigma2 = vartau,
  estimand = "PATE"
)
# Generate bias contour plot:
contour_plot(
var(weights), vartau, ipw, df_benchmark,
benchmark = TRUE, shade = TRUE,
shade_var = c("age", "prevearn"),
label_size = 4
) 
# For the external validity setting: 
data(jtpa_women)
site_name <- "NE"
df_site <- jtpa_women[which(jtpa_women$site == site_name), ]
df_else <- jtpa_women[which(jtpa_women$site != site_name), ]

# Estimate unweighted estimator:
model_dim <- estimatr::lm_robust(Y ~ T, data = df_site)
PATE <- coef(lm(Y ~ T, data = df_else))[2]
DiM <- coef(model_dim)[2]
# Generate weights using observed covariates:
df_all <- jtpa_women
df_all$S <- ifelse(jtpa_women$site == "NE", 1, 0)
model_ps <- WeightIt::weightit(
  (1 - S) ~ . - site - T - Y, 
  data = df_all, method = "ebal", estimand = "ATT"
)
weights <- model_ps$weights[df_all$S == 1]
# Estimate IPW model:
model_ipw <- estimatr::lm_robust(Y ~ T, data = df_site, weights = weights)
ipw <- coef(model_ipw)[2]
# Estimate bound for var(tau):
vartau <- var(df_site$Y[df_site$T == 1]) - var(df_site$Y[df_site$T == 0])
RV <- robustness_value(estimate = ipw, b_star = 0, sigma2 = vartau, weights = weights)

# Select weighting variables:
weighting_vars <- names(df_all)[which(!names(df_all) %in% c("site", "S", "Y", "T"))]

# Run benchmarking:
df_benchmark <- run_benchmarking(
  weighting_vars = weighting_vars,
  data = df_all[, -1],
  treatment = "T", outcome = "Y", selection = "S",
  estimate = ipw,
  RV = RV, sigma2 = vartau,
  estimand = "PATE"
)
# Generate bias contour plot:
contour_plot(
var(weights), vartau, ipw, df_benchmark,
benchmark = TRUE, shade = TRUE,
shade_var = c("age", "prevearn"),
label_size = 4
)

Helper function for creating targets from auxiliary information and formula

Description

Returns weighting targets for survey objects.

Usage

create_targets(target_design, target_formula)
create_targets(target_design, target_formula)

Arguments

target_design

A survey object

target_formula

A formula object that contains the variables to weight on

Value

Weighting target for survey objects

Examples

data(poll.data)
poll.svy = survey::svydesign(ids = ~ 1,
                     data = poll.data)

#Set up raking formula:
formula_rake = ~ age_buckets + educ + gender + race + educ * pid + bornagain

#Generate targets:
targets_rake = create_targets(poll.svy, formula_rake)

data(poll.data)
poll.svy = survey::svydesign(ids = ~ 1,
                     data = poll.data)

#Set up raking formula:
formula_rake = ~ age_buckets + educ + gender + race + educ * pid + bornagain

#Generate targets:
targets_rake = create_targets(poll.svy, formula_rake)

Estimate Bias

Description

Returns the bias based on the different parameters in the bias decomposition

Usage

estimate_bias(rho, R2, weights, sigma2)
estimate_bias(rho, R2, weights, sigma2)

Arguments

rho

Correlation between the error in the weights and outcome

R2

R2 measure for how much variation in the true weights is explained by the error term, must be bound on the range [0,1)

weights

Vector of estimated weights

sigma2

Estimated variance of the outcome (i.e., stats::var(Y) for obervational setting; stats::var(tau) for generalization setting)

Value

Estimated bias from omitting a confounder from weights

Examples

set.seed(331)
Y = rnorm(1000)
weights = rlogis(1000)
weights = weights/mean(weights)
estimate_bias(rho = 0.5, R2 = 0.5, weights = weights, sigma2 = var(Y))
set.seed(331)
Y = rnorm(1000)
weights = rlogis(1000)
weights = weights/mean(weights)
estimate_bias(rho = 0.5, R2 = 0.5, weights = weights, sigma2 = var(Y))

National Jobs Training Partnership Act (JTPA) Experimental Data

Description

A dataset containing the earnings and other demographic characteristics of individuals identified as women from the JTPA experiment, spanning 16 different experimental sites.

Usage

jtpa_women
jtpa_women

Format

A data frame with 6109 total observations, and 11 columns

site: Experimental site that the individual belonged to
Y: Earnings, in US dollars
T: Treatment assignment status
prevearn: Preivous earnings, in US dollars
age: Age of individual
married: Marital status (1 for married, 0 otherwise)
hrwage: Hourly Wage
black: Indicator for whether or not the individual is black
hispanic: Indicator for whether or not the individual is Hispanic
hsorged: Indicator for whether or not the individual received a high school degree, or GED
yrs_educ: Number of years of education completed.

...

Source

https://www.upjohn.org/data-tools/employment-research-data-center/national-jtpa-study

Synthetic Polling Data

Description

pid: party indicator, one of c('Democrat', 'Republican', 'Independent', 'Other')
educ: education level, one of c('High School or Less', 'Some college', 'College', 'Post-grad')
age: age of the individual
age_buckets: grouped age buckets, one of c('18to35', '36to50', '51to64', 'Over65')
bornagain: indicator for whether the individual identifies as born-again Christian, one of c('Yes', 'No')
gender: categorical gender, one of c('Men', 'Women')
race: categorical race, one of c('White', 'Black', 'Hispanic', 'Asian', 'Other')
vvweight_post: post-election verified voter weight for the individual

Usage

data('poll.data')
data('poll.data')

Details

A synthetically generated dataset that simulates polling data for a hypothetical support. The covariates are constructed to match commonly used demographic covariates in practice.

References

"Sensitivity analysis for survey weights," Political Analysis, 32(1) (2024), 1-16. #'

Examples

data('poll.data')
data('poll.data')

Robustness Value

Description

Returns the estimated robustness value, for a specified proportion change

Usage

robustness_value(estimate, b_star = 0, sigma2, weights)
robustness_value(estimate, b_star = 0, sigma2, weights)

Arguments

estimate

Weighted estimate

b_star

Threshold that corresponds to a substantively meaningful change to the research conclusion. For example, a value of 0 denotes that the bias from omitting a variable was sufficiently large to change the estimate to zero.

sigma2

Estimated variance of the outcome (i.e., stats::var(Y) for obervational setting; stats::var(tau) for generalization setting)

weights

Vector of estimated weights

Value

Robustness value for a specified proportion change

Examples

data(jtpa_women)
site_name <- "NE"
df_site <- jtpa_women[which(jtpa_women$site == site_name), ]
df_else <- jtpa_women[which(jtpa_women$site != site_name), ]

# Estimate unweighted estimator:
model_dim <- estimatr::lm_robust(Y ~ T, data = df_site)
PATE <- coef(lm(Y ~ T, data = df_else))[2]
DiM <- coef(model_dim)[2]
# Generate weights using observed covariates:
df_all <- jtpa_women
df_all$S <- ifelse(jtpa_women$site == "NE", 1, 0)
model_ps <- WeightIt::weightit(
  (1 - S) ~ . - site - T - Y, 
  data = df_all, method = "ebal", estimand = "ATT"
)
weights <- model_ps$weights[df_all$S == 1]
# Estimate IPW model:
model_ipw <- estimatr::lm_robust(Y ~ T, data = df_site, weights = weights)
ipw <- coef(model_ipw)[2]
# Estimate bound for var(tau):
vartau <- var(df_site$Y[df_site$T == 1]) - var(df_site$Y[df_site$T == 0])
RV <- robustness_value(estimate = ipw, b_star = 0, sigma2 = vartau, weights = weights)

print(RV)
data(jtpa_women)
site_name <- "NE"
df_site <- jtpa_women[which(jtpa_women$site == site_name), ]
df_else <- jtpa_women[which(jtpa_women$site != site_name), ]

# Estimate unweighted estimator:
model_dim <- estimatr::lm_robust(Y ~ T, data = df_site)
PATE <- coef(lm(Y ~ T, data = df_else))[2]
DiM <- coef(model_dim)[2]
# Generate weights using observed covariates:
df_all <- jtpa_women
df_all$S <- ifelse(jtpa_women$site == "NE", 1, 0)
model_ps <- WeightIt::weightit(
  (1 - S) ~ . - site - T - Y, 
  data = df_all, method = "ebal", estimand = "ATT"
)
weights <- model_ps$weights[df_all$S == 1]
# Estimate IPW model:
model_ipw <- estimatr::lm_robust(Y ~ T, data = df_site, weights = weights)
ipw <- coef(model_ipw)[2]
# Estimate bound for var(tau):
vartau <- var(df_site$Y[df_site$T == 1]) - var(df_site$Y[df_site$T == 0])
RV <- robustness_value(estimate = ipw, b_star = 0, sigma2 = vartau, weights = weights)

print(RV)

Run Formal Benchmarking

Description

Wrapper function to run formal benchmarking on a set of pre-specified covariates. Returns a data.frame containing the benchmarked parameter values, the estimated bias, MRCS, and minimum k_sigma and k_rho values for a killer confounder.

Usage

run_benchmarking(
  estimate,
  RV,
  formula = NULL,
  weights = NULL,
  pop_svy = NULL,
  sample_svy = NULL,
  Y = NULL,
  weighting_vars = NULL,
  benchmark_vars = "all",
  data = NULL,
  treatment = NULL,
  outcome = NULL,
  selection = NULL,
  population_targets = NULL,
  weighting_method = "ebal",
  weight_max = Inf,
  sigma2 = NULL,
  estimand = "ATT"
)
run_benchmarking(
  estimate,
  RV,
  formula = NULL,
  weights = NULL,
  pop_svy = NULL,
  sample_svy = NULL,
  Y = NULL,
  weighting_vars = NULL,
  benchmark_vars = "all",
  data = NULL,
  treatment = NULL,
  outcome = NULL,
  selection = NULL,
  population_targets = NULL,
  weighting_method = "ebal",
  weight_max = Inf,
  sigma2 = NULL,
  estimand = "ATT"
)

Arguments

estimate

Weighted estimate

RV

Robustness Value

formula

Raking formula for survey estimand

weights

A vector, containing the estimated survey weights

pop_svy

Survey object, containing the population the survey sample is being re-weighted to

sample_svy

Survey object, containing the survey sample being re-weighted

Y

outcome of interest (used for survey object)

weighting_vars

Vector of variables to use in the weights estimation for ATT or PATE

benchmark_vars

Vector of variables to benchmark parameters for. If benchmark_vars = 'all', benchmarking will be run across all variables included in the weights. If not set to all, benchmarking will be conducted across the covariates included in the vector.

data

A data.frame containing the observed covariates included in the weights; must include variables specified in weighting_vars

treatment

Denotes which variable is the treatment variable

outcome

Denotes which variable is the outcome variable

selection

Denotes which variable is the selection variable

population_targets

Population targets for the raking formula in surveys (optional, if not provided, will be generated from pop_svy)

weighting_method

Weighting method. Supports weighting methods from the package WeightIt.

weight_max

Maximum weight to trim at. Default set to Inf.

sigma2

If estimand = "PATE", sigma2 must specify the bound on treatment effect heterogeneity. For the other two estimands, the function will automatically calculate the sample variance across the control units, or the survey sample.

estimand

Specifies estimand; possible parameters include "ATT", "PATE", or "Survey"

Value

data.frame containing the benchmarked parameter values, the estimated bias, MRCS, and minimum k_sigma and k_rho values for a killer confounder for the set of pre-specified covariates.

Examples

# For the external validity setting: 
data(jtpa_women)
site_name <- "NE"
df_site <- jtpa_women[which(jtpa_women$site == site_name), ]
df_else <- jtpa_women[which(jtpa_women$site != site_name), ]

# Estimate unweighted estimator:
model_dim <- estimatr::lm_robust(Y ~ T, data = df_site)
PATE <- coef(lm(Y ~ T, data = df_else))[2]
DiM <- coef(model_dim)[2]
# Generate weights using observed covariates:
df_all <- jtpa_women
df_all$S <- ifelse(jtpa_women$site == "NE", 1, 0)
model_ps <- WeightIt::weightit(
  (1 - S) ~ . - site - T - Y, 
  data = df_all, method = "ebal", estimand = "ATT"
)
weights <- model_ps$weights[df_all$S == 1]
# Estimate IPW model:
model_ipw <- estimatr::lm_robust(Y ~ T, data = df_site, weights = weights)
ipw <- coef(model_ipw)[2]
# Estimate bound for var(tau):
vartau <- var(df_site$Y[df_site$T == 1]) - var(df_site$Y[df_site$T == 0])
RV <- robustness_value(estimate = ipw, b_star = 0, sigma2 = vartau, weights = weights)

# Select weighting variables:
weighting_vars <- names(df_all)[which(!names(df_all) %in% c("site", "S", "Y", "T"))]

# Run benchmarking:
df_benchmark <- run_benchmarking(
  weighting_vars = weighting_vars,
  data = df_all[, -1],
  treatment = "T", outcome = "Y", selection = "S",
  estimate = ipw,
  RV = RV, sigma2 = vartau,
  estimand = "PATE"
)

print(df_benchmark)

# For the external validity setting: 
data(jtpa_women)
site_name <- "NE"
df_site <- jtpa_women[which(jtpa_women$site == site_name), ]
df_else <- jtpa_women[which(jtpa_women$site != site_name), ]

# Estimate unweighted estimator:
model_dim <- estimatr::lm_robust(Y ~ T, data = df_site)
PATE <- coef(lm(Y ~ T, data = df_else))[2]
DiM <- coef(model_dim)[2]
# Generate weights using observed covariates:
df_all <- jtpa_women
df_all$S <- ifelse(jtpa_women$site == "NE", 1, 0)
model_ps <- WeightIt::weightit(
  (1 - S) ~ . - site - T - Y, 
  data = df_all, method = "ebal", estimand = "ATT"
)
weights <- model_ps$weights[df_all$S == 1]
# Estimate IPW model:
model_ipw <- estimatr::lm_robust(Y ~ T, data = df_site, weights = weights)
ipw <- coef(model_ipw)[2]
# Estimate bound for var(tau):
vartau <- var(df_site$Y[df_site$T == 1]) - var(df_site$Y[df_site$T == 0])
RV <- robustness_value(estimate = ipw, b_star = 0, sigma2 = vartau, weights = weights)

# Select weighting variables:
weighting_vars <- names(df_all)[which(!names(df_all) %in% c("site", "S", "Y", "T"))]

# Run benchmarking:
df_benchmark <- run_benchmarking(
  weighting_vars = weighting_vars,
  data = df_all[, -1],
  treatment = "T", outcome = "Y", selection = "S",
  estimate = ipw,
  RV = RV, sigma2 = vartau,
  estimand = "PATE"
)

print(df_benchmark)

Sensitivity Summary

Description

Returns a data.frame or Kable table with summary measures of sensitivity

Usage

summarize_sensitivity(
  weights = NULL,
  Y = NULL,
  Z = NULL,
  b_star = 0,
  estimate = NULL,
  SE = NULL,
  unweighted = NULL,
  sigma2 = NULL,
  estimand = "ATT",
  pretty = FALSE,
  svy_srs = NULL,
  svy_wt = NULL,
  sig.fig = 2
)
summarize_sensitivity(
  weights = NULL,
  Y = NULL,
  Z = NULL,
  b_star = 0,
  estimate = NULL,
  SE = NULL,
  unweighted = NULL,
  sigma2 = NULL,
  estimand = "ATT",
  pretty = FALSE,
  svy_srs = NULL,
  svy_wt = NULL,
  sig.fig = 2
)

Arguments

weights

Vector of estimated weights

Y

Outcome of interest

Z

Treatment assignment (Not needed for settings when users are analyzing surveys)

b_star

Killer confounder threshold. If not specified, will be automatically set to 0.

estimate

(Optional) Weighted point estimate. If not specified, function will automatically generate the weighted estimator, given the inputs in Y and weights

SE

(Optional) Standard error associated with the weighted point estimate

unweighted

(Optional) Unweighted point estimate.

sigma2

(Optional) Variance of outcomes or individual-level treatment effect in PATE case. In the case of a PATE estimator, if not specified, function will automatically estimate an upper bound for the variance of the individual-level treatment effect.

estimand

Specifies estimand; possible parameters include "ATT", "PATE", or "Survey"

pretty

If set to TRUE, will return a Kable table. If set to FALSE, will return a data.frame.

svy_srs

Unweighted svymean object

svy_wt

Weighted svymean object

sig.fig

Significant figures to round the output to (default set to 2)

Value

Sensitivity summary

Examples

data(jtpa_women)
site_name <- "NE"
df_site <- jtpa_women[which(jtpa_women$site == site_name), ]
df_else <- jtpa_women[which(jtpa_women$site != site_name), ]

# Estimate unweighted estimator:
model_dim <- estimatr::lm_robust(Y ~ T, data = df_site)
PATE <- coef(lm(Y ~ T, data = df_else))[2]
DiM <- coef(model_dim)[2]
# Generate weights using observed covariates:
df_all <- jtpa_women
df_all$S <- ifelse(jtpa_women$site == "NE", 1, 0)
model_ps <- WeightIt::weightit(
  (1 - S) ~ . - site - T - Y, 
  data = df_all, method = "ebal", estimand = "ATT"
)
weights <- model_ps$weights[df_all$S == 1]
# Estimate IPW model:
model_ipw <- estimatr::lm_robust(Y ~ T, data = df_site, weights = weights)
ipw <- coef(model_ipw)[2]
# Estimate bound for var(tau):
vartau <- var(df_site$Y[df_site$T == 1]) - var(df_site$Y[df_site$T == 0])
summarize_sensitivity(weights = weights, 
Y = df_site$Y, 
Z = df_site$T, 
sigma2 = vartau,
 estimand = "PATE")
data(jtpa_women)
site_name <- "NE"
df_site <- jtpa_women[which(jtpa_women$site == site_name), ]
df_else <- jtpa_women[which(jtpa_women$site != site_name), ]

# Estimate unweighted estimator:
model_dim <- estimatr::lm_robust(Y ~ T, data = df_site)
PATE <- coef(lm(Y ~ T, data = df_else))[2]
DiM <- coef(model_dim)[2]
# Generate weights using observed covariates:
df_all <- jtpa_women
df_all$S <- ifelse(jtpa_women$site == "NE", 1, 0)
model_ps <- WeightIt::weightit(
  (1 - S) ~ . - site - T - Y, 
  data = df_all, method = "ebal", estimand = "ATT"
)
weights <- model_ps$weights[df_all$S == 1]
# Estimate IPW model:
model_ipw <- estimatr::lm_robust(Y ~ T, data = df_site, weights = weights)
ipw <- coef(model_ipw)[2]
# Estimate bound for var(tau):
vartau <- var(df_site$Y[df_site$T == 1]) - var(df_site$Y[df_site$T == 0])
summarize_sensitivity(weights = weights, 
Y = df_site$Y, 
Z = df_site$T, 
sigma2 = vartau,
 estimand = "PATE")

Sensitivity Summary (for survey weights)

Description

Returns a data.frame or Kable table with summary measures of sensitivity; helper function for main summary function, and allows users to directly input a survey object and a design object

Usage

summarize_sensitivity_survey(svy_srs, svy_wt, weights, varY, b_star = 0)
summarize_sensitivity_survey(svy_srs, svy_wt, weights, varY, b_star = 0)

Arguments

svy_srs

Survey object, containing the unweighted survey

svy_wt

Survey object, containing the weighted survey

weights

A vector, containing the estimated survey weights

varY

variance of the outcome

b_star

Killer confounder threshold, default set to be zero

Value

Sensitivity data.frame or latex table with summary measures of sensitivity

Examples

data(poll.data)
poll_srs <- survey::svydesign(ids = ~ 1, data = poll.data)
pop_targets = c(1, 0.212, 0.264, 0.236, 0.310, 
                0.114, 0.360, 0.528, 0.114, 
                0.021, 0.034, 0.805, 
                0.266, 0.075, 0.312, 0.349)
names(pop_targets) = c("(Intercept)",
                       "age_buckets36to50",
                       "age_buckets51to64",
                       "age_bucketsOver65",
                       "educHigh School or Less",
                       "educPost-grad",
                       "educSome college",
                       "genderWomen", 
                       "raceBlack",
                       "raceHispanic",
                       "raceOther",
                       "raceWhite", 
                       "pidIndependent", "pidOther", 
                       "pidRepublican", "bornagainYes")
#Set up raking formula:
formula_rake <- ~ age_buckets + educ + gender + race + pid + bornagain

#PERFORM RAKING:
model_rake <- survey::calibrate(
  design = poll_srs,
  formula = formula_rake,
  population = pop_targets,
  calfun = "raking",
  force = TRUE
)


rake_results <- survey::svydesign( ~ 1, data = poll.data, weights = stats::weights(model_rake))
#Estimate from raking results:
weights = stats::weights(rake_results) * nrow(model_rake)

unweighted_estimate = survey::svymean(~ Y, poll_srs, na.rm = TRUE)
weighted_estimate = survey::svymean(~ Y, model_rake, na.rm = TRUE)
summarize_sensitivity(estimand = 'Survey',
Y = poll.data$Y,
weights = weights,
svy_srs = unweighted_estimate, 
svy_wt = weighted_estimate,
b_star = 0.5)
data(poll.data)
poll_srs <- survey::svydesign(ids = ~ 1, data = poll.data)
pop_targets = c(1, 0.212, 0.264, 0.236, 0.310, 
                0.114, 0.360, 0.528, 0.114, 
                0.021, 0.034, 0.805, 
                0.266, 0.075, 0.312, 0.349)
names(pop_targets) = c("(Intercept)",
                       "age_buckets36to50",
                       "age_buckets51to64",
                       "age_bucketsOver65",
                       "educHigh School or Less",
                       "educPost-grad",
                       "educSome college",
                       "genderWomen", 
                       "raceBlack",
                       "raceHispanic",
                       "raceOther",
                       "raceWhite", 
                       "pidIndependent", "pidOther", 
                       "pidRepublican", "bornagainYes")
#Set up raking formula:
formula_rake <- ~ age_buckets + educ + gender + race + pid + bornagain

#PERFORM RAKING:
model_rake <- survey::calibrate(
  design = poll_srs,
  formula = formula_rake,
  population = pop_targets,
  calfun = "raking",
  force = TRUE
)


rake_results <- survey::svydesign( ~ 1, data = poll.data, weights = stats::weights(model_rake))
#Estimate from raking results:
weights = stats::weights(rake_results) * nrow(model_rake)

unweighted_estimate = survey::svymean(~ Y, poll_srs, na.rm = TRUE)
weighted_estimate = survey::svymean(~ Y, model_rake, na.rm = TRUE)
summarize_sensitivity(estimand = 'Survey',
Y = poll.data$Y,
weights = weights,
svy_srs = unweighted_estimate, 
svy_wt = weighted_estimate,
b_star = 0.5)

Package 'senseweight'

Help Index

Benchmark (for survey weights)

Description

Usage

Arguments

Value

Examples

Bias Contour Plots

Description

Usage

Arguments

Value

Examples

Helper function for creating targets from auxiliary information and formula

Description

Usage

Arguments

Value

Examples

Estimate Bias

Description

Usage

Arguments

Value

Examples

National Jobs Training Partnership Act (JTPA) Experimental Data

Description

Usage

Format

Source

Synthetic Polling Data

Description

Usage

Details

References

Examples

Robustness Value

Description

Usage

Arguments

Value

Examples

Run Formal Benchmarking

Description

Usage

Arguments

Value

Examples

Sensitivity Summary

Description

Usage

Arguments

Value

Examples

Sensitivity Summary (for survey weights)

Description

Usage

Arguments

Value

Examples