### Install and Manage Causal Inference Task View

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Installs, updates, or reads the metadata for the CausalInference task view using the ctv package. Ensure the ctv package is installed first.

```r
# Install the ctv package
install.packages("ctv")
```

```r
# Install all packages in the CausalInference task view
ctv::install.views("CausalInference")
```

```r
# Update already-installed task view packages
ctv::update.views("CausalInference")
```

```r
# Read the task view metadata locally (for contributors)
ctv::read.ctv("CausalInference.md")
#> $name
#> [1] "CausalInference"
#> $packagelist
#>         package priority
#> 1  MatchIt          core
#> 2  WeightIt         core
#> ...

```

--------------------------------

### Get Synthetic Control Significance

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Retrieves and displays significance test results for the synthetic control analysis, including p-values and z-scores.

```R
library(tidysynth)

synth_out |> grab_significance()
#>   unit_name  type  fishers_exact_pvalue  z_score
#>   California Treated          0.0370    -2.41
```

--------------------------------

### Create and Plot a Policy Tree

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Uses the policytree package to create a policy tree from data and then plots the resulting tree.

```R
library(policytree)
pt <- policy_tree(X, cf, depth = 2)
plot(pt)
```

--------------------------------

### Export Causal Packages to File

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Exports a unique list of causal package names to a text file for manual review. Displays the count of candidate packages found.

```R
write.table(unique(causal_pkgs$Package),
            file      = "data/causal_packages.txt",
            row.names = FALSE, col.names = FALSE, quote = FALSE)
message(n_distinct(causal_pkgs$Package), " candidate packages found")
```

--------------------------------

### Build Synthetic Control with tidysynth

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Constructs a synthetic control group using the tidysynth package. This involves generating predictors and weights based on specified time windows and optimization.

```R
library(tidysynth)

data("smoking", package = "tidysynth")   # California Prop 99 tobacco data

synth_out <- smoking |>
  synthetic_control(outcome    = cigsale,
                    unit       = unit,
                    time       = year,
                    i_unit     = "California",
                    i_time     = 1988,
                    generate_placebos = TRUE) |>
  generate_predictor(time_window = 1980:1988,
                     ln_income     = mean(lnincome),
                     retail_price  = mean(retprice),
                     beer_sales    = mean(beer)) |> 
  generate_predictor(time_window = 1984:1988,
                     cigsale_84_88 = mean(cigsale)) |>
  generate_weights(optimization_window = 1970:1988) |>
  generate_control()
```

--------------------------------

### Learn Causal Structure with PC Algorithm

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Applies the PC algorithm from the pcalg package to learn a causal structure from observational data. Requires a suffStat object containing correlation matrix and sample size.

```R
library(pcalg)
library(dagitty)
library(ggdag)

# 1. PC algorithm for causal structure learning from observational data
data("gmG8", package = "pcalg")     # Gaussian data, 8 variables
suffStat <- list(C = cor(gmG8$x), n = nrow(gmG8$x))
pc_fit   <- pc(suffStat,
               indepTest = gaussCItest,
               alpha     = 0.05,
               labels    = colnames(gmG8$x),
               verbose   = FALSE)
plot(pc_fit, main = "PC Algorithm CPDAG")
```

--------------------------------

### Estimate ATE using Entropy Balancing Weights

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Estimates ATE using entropy balancing weights with the WeightIt package and fits a weighted linear model. Requires the WeightIt package.

```r
# 2. Entropy balancing weights
w_out <- weightit(treat ~ age + educ + race + re74 + re75,
                  data   = lalonde,
                  method = "ebal",
                  estimand = "ATE")
fit_w  <- lm(re78 ~ treat, data = lalonde, weights = w_out$weights)
summary(fit_w)

```

--------------------------------

### Estimate Average Treatment Effect (ATE) using Matching

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Performs propensity score matching using the MatchIt package and estimates the ATE by fitting a linear model on the matched data. Requires the MatchIt and marginaleffects packages.

```r
library(MatchIt)
library(WeightIt)
library(marginaleffects)

data("lalonde", package = "MatchIt")

# 1. Propensity score matching
m_out <- matchit(treat ~ age + educ + race + re74 + re75,
                 data   = lalonde,
                 method = "nearest",
                 ratio  = 1)
m_data <- match.data(m_out)
fit_m  <- lm(re78 ~ treat + age + educ + race + re74 + re75,
             data    = m_data,
             weights = weights)
avg_comparisons(fit_m, variables = "treat")
#>  Estimate Std. Error     z Pr(>|z|)   S  2.5 % 97.5 %
#>      1205        578  2.08   0.0372 4.75   71   2338

```

--------------------------------

### Discover Causal Inference CRAN Packages

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Automates the discovery of relevant CRAN packages for causal inference by text-mining package descriptions and comparing them against a list of keywords and existing task view packages.

```R
library(tidytext)
library(tidyverse)

# Keywords signalling causal-inference relevance
mda_words <- c("causal", "causal inference", "treatment effect", "ATE", "HTE",
               "propensity score", "potential outcomes", "RCT",
               "instrumental variable", "IPW", "do-calculus")

# Read currently listed packages from the task view source
listed_packages <- ctv::read.ctv("CausalInference.md")$packagelist[, 1]

# Download full CRAN package database
cran_db  <- tools::CRAN_package_db()
cran_tbl <- tibble::as_tibble(cran_db[, -65]) |>
              select(Package, Description)

# Tokenise descriptions and filter by keywords
data("stop_words")
causal_pkgs <- cran_tbl |>
  unnest_tokens(word, Description) |>
  anti_join(stop_words, by = "word") |>
  group_by(Package) |>
  filter(tolower(word) %in% mda_words) |>
  ungroup()

# New candidates not yet in the task view
new_candidates <- causal_pkgs |>
  filter(!Package %in% listed_packages) |> 
  distinct(Package)
```

--------------------------------

### Doubly Robust Estimation with TMLE and AIPW

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Demonstrates TMLE and AIPW for doubly robust estimation. Requires outcome and propensity model specification. TMLE uses Super Learner for flexible modeling.

```r
library(tmle)

set.seed(1)
n  <- 500
W  <- data.frame(W1 = rnorm(n), W2 = rbinom(n, 1, 0.4))
A  <- rbinom(n, 1, plogis(0.3 * W$W1 - 0.2 * W$W2))
Y  <- rnorm(n, mean = 2 * A + W$W1 - W$W2, sd = 1)

# TMLE with Super Learner
result <- tmle(Y = Y, A = A, W = W,
               family = "gaussian",
               g.SL.library = c("SL.glm", "SL.ranger"),
               Q.SL.library = c("SL.glm", "SL.ranger"))
summary(result)
#> Additive Effect
#>  Parameter Estimate:  1.98
#>  Estimated Variance:  0.0042
#>
#>             p-value:  <2e-16
#>
#>   95% Conf Interval: (1.85, 2.11)
```

```r
library(AIPW)
aipw <- AIPW$new(Y = Y, A = A, W = W,
                 Q.SL.library = c("SL.glm","SL.ranger"),
                 g.SL.library = c("SL.glm","SL.ranger"),
                 k_split = 3)
aipw$fit()$summary()
```

--------------------------------

### Identify Adjustment Set using DAGitty

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Uses the dagitty package to define a Directed Acyclic Graph (DAG) and identify necessary adjustment sets for estimating causal effects.

```R
library(pcalg)
library(dagitty)
library(ggdag)

# 2. Identify adjustment set using DAGitty
dag <- dagitty("dag {
  X -> M -> Y
  X -> Y
  U -> M
  U -> Y
}")
adjustmentSets(dag, exposure = "X", outcome = "Y")
#> { }   (no adjustment needed for total effect X->Y, but M is a collider)
```

--------------------------------

### Analyze Randomized Controlled Trials (RCTs)

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Provides tools for simulating RCT data and performing intent-to-treat analysis using linear models. Also includes power analysis for multi-level RCTs.

```r
library(experiment)

# Simulate a simple two-arm RCT
set.seed(42)
n <- 200
data_rct <- data.frame(
  Y  = rnorm(n, mean = 5, sd = 2),
  Z  = rbinom(n, 1, 0.5),           # treatment assignment
  X1 = rnorm(n),                    # covariate
  X2 = rbinom(n, 1, 0.4)
)

# Intent-to-treat analysis
itt <- lm(Y ~ Z + X1 + X2, data = data_rct)
summary(itt)
#> Coefficients:
#>             Estimate Std. Error t value Pr(>|t|)
#> (Intercept)  4.9821     0.2213  22.51   <2e-16 ***
#> Z            0.1234     0.2819   0.44    0.662

```

```r
# Power analysis for multi-level RCT
library(cosa)
cosa.Bird2016(
  es = 0.25,        # target effect size
  n = c(20, 10),    # students per classroom, classrooms per school
  p = 0.5           # proportion treated
)

```

--------------------------------

### Estimate ATE using Double Machine Learning

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Applies Double/debiased Machine Learning (DML) for ATE estimation using the DoubleML package with ranger learners. Requires DoubleML, mlr3, and mlr3learners packages.

```r
# 3. Double/debiased machine learning
library(DoubleML)
library(mlr3)
library(mlr3learners)
dml_data <- DoubleMLData$new(lalonde,
                              y_col = "re78",
                              d_cols = "treat",
                              x_cols = c("age","educ","re74","re75"))
lrn_rf <- lrn("regr.ranger", num.trees = 200)
dml_plr <- DoubleMLPLR$new(dml_data, ml_l = lrn_rf, ml_m = lrn_rf)
dml_plr$fit()
dml_plr$summary()
#> Estimates and significance testing of the effect of target variables
#>       Estimate. Std. Error t value Pr(>|t|)
#> treat    1423.0      517.3   2.751  0.00594 **

```

--------------------------------

### Visualize DAG and Adjustment Set with ggdag

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Visualizes a DAG and its identified adjustment set using ggplot2 via the ggdag package.

```R
library(pcalg)
library(dagitty)
library(ggdag)

# 3. Visualise DAG in ggplot2
ggdag_adjustment_set(
  tidy_dagitty(dag),
  exposure = "X", outcome = "Y"
) + theme_dag()
```

--------------------------------

### Regression Discontinuity Design (RDD)

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Applies RDD estimation with optimal bandwidth selection and robust bias-corrected confidence intervals. Includes density tests for manipulation.

```r
library(rdrobust)

data("rdrobust_RDsenate", package = "rdrobust")
# x = margin of victory, y = vote share in next election

rdd <- rdrobust(y = rdrobust_RDsenate$vote,
                x = rdrobust_RDsenate$margin,
                kernel = "triangular",
                bwselect = "mserd")
summary(rdd)
#> RD Estimates:
#>           Coef. Std. Err.     z      P>|z|   [ 95% C.I. ]
#> Conventional  6.97  1.43   4.87  <0.001  [ 4.16,  9.77]
#> Bias-Corrected 7.00  1.43   4.89  <0.001  [ 4.19,  9.81]
#> Robust         7.00  1.74   4.02  <0.001  [ 3.59, 10.41]
```

```r
rddensity::rddensity(rdrobust_RDsenate$margin) |> summary()
```

```r
rdplot(y = rdrobust_RDsenate$vote, x = rdrobust_RDsenate$margin,
       title = "Senate Incumbency Advantage")
```

--------------------------------

### Causal Forests for Heterogeneous Treatment Effects

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Estimates heterogeneous treatment effects using causal forests. Requires specifying covariates (X), outcomes (Y), and treatment assignment (W). Supports ATE, CATE, and heterogeneity testing.

```r
library(grf)

data("lalonde", package = "MatchIt")
X  <- model.matrix(~ age + educ + re74 + re75 - 1, data = lalonde)
Y  <- lalonde$re78
W  <- lalonde$treat

# Causal forest
cf <- causal_forest(X = X, Y = Y, W = W,
                    num.trees = 2000,
                    tune.parameters = "all")

# Average treatment effect
average_treatment_effect(cf, target.sample = "all")
#>   estimate    std.err
#>    1462.41     540.63

# Conditional average treatment effects (CATEs)
cates <- predict(cf)$predictions

# Best linear projection onto covariates
best_linear_projection(cf, A = X)

# Test for heterogeneity
test_calibration(cf)
#>                        Estimate Std. Error t value  Pr(>|t|)
#> mean.forest.prediction   1.0022     0.4519   2.218  0.02716 *
#> differential.forest.pred 0.9718     0.6105   1.592  0.11188
```

--------------------------------

### Plot Synthetic Control Placebos

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Generates and plots placebo tests for the synthetic control analysis, which involves running the synthetic control method on different units or time periods.

```R
library(tidysynth)

synth_out |> plot_placebos()   # permutation-based inference
```

--------------------------------

### Perform Causal Mediation Analysis

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Conducts causal mediation analysis using the mediation package, estimating Average Causal Mediation Effects (ACME) and Average Direct Effects (ADE). Requires fitting mediator and outcome models first.

```R
library(mediation)

data("jobs", package = "mediation")

# Fit mediator and outcome models
model.m <- lm(job_seek ~ treat + depress1 + econ_hard + sex + age,
              data = jobs)
model.y <- lm(depress2 ~ treat + job_seek + depress1 + econ_hard + sex + age,
              data = jobs)

# Causal mediation analysis (ACME + ADE)
set.seed(2014)
med_out <- mediate(model.m, model.y,
                   treat = "treat", mediator = "job_seek",
                   boot = TRUE, sims = 1000)
summary(med_out)
#>                Estimate 95% CI Lower 95% CI Upper p-value
#> ACME (control)  -0.0142      -0.0307         0.00    0.08
#> ACME (treated)  -0.0142      -0.0307         0.00    0.08
#> ADE (control)   -0.0439      -0.1208         0.03    0.26
#> Total Effect    -0.0581      -0.1371         0.02    0.17
#> Prop. Mediated   0.2443      -0.5843         1.57    0.17
```

--------------------------------

### Plot Synthetic Control Differences

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Visualizes the difference between the outcome of the treated unit and its synthetic control over time.

```R
library(tidysynth)

synth_out |> plot_differences()
```

--------------------------------

### Difference-in-Differences with Staggered Adoption

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Implements Callaway & Sant'Anna (2021) and Sun & Abraham estimators for DiD with staggered treatment adoption. Requires specifying outcome, treatment timing, and covariates.

```r
library(did)

# Callaway & Sant'Anna (2021) estimator with staggered treatment
data("mpdta", package = "did")
# mpdta: county-level minimum wage data, first.treat = year of treatment
out <- att_gt(yname       = "lemp",
              gname       = "first.treat",
              idname      = "countyreal",
              tname       = "year",
              xformla     = ~lpop,
              data        = mpdta,
              est_method  = "dr")         # doubly robust

# Aggregate to simple ATT
simple_att <- aggte(out, type = "simple")
summary(simple_att)
#> ATT    SE  [ 95%  Conf. Int.]
#> -0.0323 0.0124  -0.0565  -0.0080

# Event-study plot
ggdid(aggte(out, type = "dynamic"), ylim = c(-0.15, 0.10))
```

```r
library(fixest)
feols(lemp ~ sunab(first.treat, year) + lpop | countyreal + year,
      data = mpdta) |> iplot()
```

--------------------------------

### Sensitivity Analysis for Unmeasured Confounding

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Performs a sensitivity analysis for unmeasured confounding in mediation analysis using the medsens function from the mediation package.

```R
library(mediation)

sens_out <- medsens(med_out, rho.by = 0.05, effect.type = "indirect")
plot(sens_out, main = "Sensitivity: ACME")
```

--------------------------------

### Instrumental Variables Regression

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Performs 2SLS and LIML estimation for instrumental variables. The first formula specifies endogenous and exogenous covariates, while the second specifies instruments.

```r
library(ivreg)

# Classic Card (1995) returns-to-education IV
data("CigarettesSW", package = "AER")

# Two-stage least squares: proximity to college as instrument for education
iv_fit <- ivreg(log(packs) ~ log(price) + log(income) |
                              log(rincome) + tdiff,
                data = CigarettesSW)
summary(iv_fit, diagnostics = TRUE)
#> Diagnostic tests:
#>
#>                   df1 df2 statistic p-value
#>  Weak instruments    1  44    244.73  <2e-16 ***
#>  Wu-Hausman          1  43      5.12  0.0286 *
#>  Sargan              0  NA        NA      NA
```

```r
library(fixest)
feiv(re78 ~ 1 | age + educ | treat ~ nearc4,
     data = MatchIt::lalonde)
```

--------------------------------

### Plot Synthetic Control Trends

Source: https://context7.com/cran-task-views/causalinference/llms.txt

Plots the trends of the outcome variable for the treated unit and its synthetic control.

```R
library(tidysynth)

synth_out |> plot_trends()
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.