triplediff

http://marcelortiz.com/triplediff/
Admin
triplediff is an R package for computing average treatment effects in Triple Differences (DDD)...
Tokens:6,055
Snippets:57
Trust Score:6.3
Update:1 month ago
Show doc for...
Tokens:
Raw
### Install and Load triplediff

Source: http://marcelortiz.com/triplediff/index.html

Install the development version from GitHub and load the library into the R environment.

```R
# install.packages("devtools")
devtools::install_github("marcelortizv/triplediff")

library(triplediff)
```

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

### Triple-Differences Aggregation Example

Source: http://marcelortiz.com/triplediff/reference/agg_ddd.html

Example demonstrating data generation, model estimation, and simple aggregation of treatment effects.

```R
#----------------------------------------------------------
# Triple Diff with multiple time periods
#----------------------------------------------------------

data <- gen_dgp_mult_periods(size = 500, dgp_type = 1)[["data"]]

out <- ddd(yname = "y", tname = "time", idname = "id",
            gname = "state", pname = "partition", xformla = ~cov1 + cov2 + cov3 + cov4,
            data = data, control_group = "nevertreated", base_period = "varying",
            est_method = "dr")
# Simple aggregation
agg_ddd(out, type = "simple", alpha = 0.10)
```

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

### Execute DDD estimation

Source: http://marcelortiz.com/triplediff/reference/ddd.html

Example call to the ddd function specifying outcome, time, id, group, and partition variables along with covariates and estimation method.

```R
ddd(yname = "y", tname = "time", idname = "id",
     gname = "state", pname = "partition", xformla = ~cov1 + cov2 + cov3 + cov4,
     data = data, control_group = "nevertreated", base_period = "varying",
     est_method = "dr")
```

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

### Simulate Data for Two-Period DDD

Source: http://marcelortiz.com/triplediff/index.html

Generate a synthetic dataset for a two-period DDD setup using the built-in gen_dgp_2periods function.

```R
set.seed(1234) # Set seed for reproducibility
# Simulate data for a two-periods DDD setup
df <- gen_dgp_2periods(
  size = 5000, # Number of units
  dgp_type = 1 # Type of DGP design
)$data

head(df)
```

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

### Simulate Multiple-Period DDD Data

Source: http://marcelortiz.com/triplediff/index.html

Generates a dataset for a multiple-period DDD setup with staggered treatment adoption using the gen_dgp_mult_periods function.

```R
set.seed(1234) # Set seed for reproducibility
# Simulate data for a multiple-period DDD setup with staggered treatment adoption
data <- gen_dgp_mult_periods(size = 500, dgp_type = 1)$data

head(data)
```

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

### Generate Data for Multiple Time Periods

Source: http://marcelortiz.com/triplediff/reference/ddd.html

Initializes a dataset for a triple difference analysis spanning multiple time periods.

```R
#----------------------------------------------------------
# Triple Diff with multiple time periods
#----------------------------------------------------------
data <- gen_dgp_mult_periods(size = 1000, dgp_type = 1)[["data"]]
```

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

### Initialize ddd estimator

Source: http://marcelortiz.com/triplediff/reference/ddd.html

The primary function call for computing Doubly Robust DDD estimators. Ensure all required column names for outcome, time, group, and partition variables are provided.

```R
ddd(
  yname,
  tname,
  idname = NULL,
  gname,
  pname,
  xformla,
  data,
  control_group = NULL,
  base_period = NULL,
  est_method = "dr",
  panel = TRUE,
  allow_unbalanced_panel = FALSE,
  weightsname = NULL,
  boot = FALSE,
  nboot = NULL,
  cluster = NULL,
  cband = FALSE,
  alpha = 0.05,
  use_parallel = FALSE,
  cores = 1,
  inffunc = FALSE,
  skip_data_checks = FALSE
)
```

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

### ddd(yname, tname, idname, gname, pname, xformla, data, ...)

Source: http://marcelortiz.com/triplediff/reference/ddd.html

Computes the Doubly Robust DDD estimators for the ATT. This is the primary function for the package.

```APIDOC
## ddd

### Description
Computes the Doubly Robust DDD estimators for the ATT with balanced panel data, supporting covariates and multiple time periods.

### Parameters
- **yname** (string) - Required - The name of the outcome variable.
- **tname** (string) - Required - The name of the column containing the time periods.
- **idname** (string) - Optional - The name of the column containing the unit id.
- **gname** (string) - Required - The name of the column containing the first period when a particular observation is treated.
- **pname** (string) - Optional - The name of the column containing the partition variable.
- **xformla** (formula) - Optional - The formula for the covariates to be included in the model (default: ~1).
- **data** (data.frame/data.table) - Required - The data to be used.
- **control_group** (string) - Optional - The control group to be used (default: "notyettreated").
- **base_period** (string) - Optional - Choose between "varying" or "universal" base period.
- **est_method** (string) - Optional - The estimation method: "dr", "reg", or "ipw" (default: "dr").
- **panel** (boolean) - Optional - Whether data is treated as panel data (default: TRUE).
- **allow_unbalanced_panel** (boolean) - Optional - Whether to allow unbalanced panel data (default: FALSE).
- **weightsname** (string) - Optional - The name of the column containing the weights.
- **boot** (boolean) - Optional - Whether to compute standard errors using multiplier bootstrap (default: FALSE).
- **nboot** (integer) - Optional - Number of bootstrap samples (default: 999 if boot=TRUE).
- **cluster** (string) - Optional - The name of the variable to be used for clustering.
- **cband** (boolean) - Optional - Whether to compute a uniform confidence band (default: FALSE).
- **alpha** (numeric) - Optional - Significance level for confidence intervals (default: 0.05).
- **use_parallel** (boolean) - Optional - Whether to use parallel processing (default: FALSE).
- **cores** (integer) - Optional - Number of cores for parallel processing (default: 1).
- **inffunc** (boolean) - Optional - Whether to return the influence function (default: FALSE).
- **skip_data_checks** (boolean) - Optional - Whether to skip data checks (default: FALSE).
```

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

### Prepare Aggregated Data for Visualization

Source: http://marcelortiz.com/triplediff/index.html

Extract event-study estimates and confidence intervals into a tibble for plotting with ggplot2.

```R
es_df <- with(es_e$aggte_ddd, {
    tibble::tibble(period   = egt,
                   estimate = att.egt,
                   ci_lower = estimate - crit.val.egt * se.egt,
                   ci_upper = estimate + crit.val.egt * se.egt)
  })


all_per <- sort(unique(es_df$period))
```

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

### gen_dgp_mult_periods()

Source: http://marcelortiz.com/triplediff/reference/index.html

Generate panel data with staggered treatment adoption (three periods).

```APIDOC
## gen_dgp_mult_periods()

### Description
Generate panel data with staggered treatment adoption (three periods).
```

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

### ddd()

Source: http://marcelortiz.com/triplediff/reference/index.html

Doubly Robust DDD estimators for the group-time average treatment effects.

```APIDOC
## ddd()

### Description
Doubly Robust DDD estimators for the group-time average treatment effects.
```

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

### Usage of gen_dgp_2periods

Source: http://marcelortiz.com/triplediff/reference/gen_dgp_2periods.html

Call this function with the desired sample size and a DGP type integer to generate a list containing the panel data and theoretical parameters.

```R
gen_dgp_2periods(size, dgp_type)
```

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

### Usage of gen_dgp_mult_periods

Source: http://marcelortiz.com/triplediff/reference/gen_dgp_mult_periods.html

Call this function to generate a simulated dataset with specified unit size and model configuration.

```R
gen_dgp_mult_periods(size, dgp_type = 1, include_covariates = TRUE)
```

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

### Function Usage for agg_ddd

Source: http://marcelortiz.com/triplediff/reference/agg_ddd.html

Standard signature for the agg_ddd function.

```R
agg_ddd(
  ddd_obj,
  type = "eventstudy",
  cluster = NULL,
  balance_e = NULL,
  min_e = -Inf,
  max_e = Inf,
  na.rm = FALSE,
  boot = NULL,
  nboot = NULL,
  cband = NULL,
  alpha = 0.05
)
```

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

### BibTeX Citation for R Package

Source: http://marcelortiz.com/triplediff/authors.html

Use this BibTeX entry to cite the triplediff R package version 0.2.4.

```BibTeX
@Misc{,
  title = {triplediff: Triple Differences Estimators},
  author = {Marcelo Ortiz-Villavicencio and Pedro H.C. Sant'Anna},
  year = {2025},
  note = {R package version 0.2.4},
  url = {http://marcelortiz.com/triplediff/},
}
```

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

### Aggregate DDD results by event study

Source: http://marcelortiz.com/triplediff/reference/agg_ddd.html

Performs dynamic aggregation of ATT estimates based on event time.

```R
agg_ddd(out, type = "eventstudy", alpha = 0.10)
```

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

### Aggregate DDD results by calendar time

Source: http://marcelortiz.com/triplediff/reference/agg_ddd.html

Performs aggregation of ATT estimates based on calendar time.

```R
agg_ddd(out, type = "calendar", alpha = 0.10)
```

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

### Visualize DDD estimates with ggplot2

Source: http://marcelortiz.com/triplediff/index.html

Creates an error bar plot for DDD estimates with custom theme and axis formatting.

```R
p <- ggplot(es_df, aes(period, estimate)) +
    geom_errorbar(aes(ymin = ci_lower, ymax = ci_upper),
                  width = .15,
                  position = position_dodge(width = .3)) +
    geom_point(size = 2.2,
               position = position_dodge(width = .3)) +
    geom_hline(yintercept = 0, linewidth = .3) +
    geom_vline(xintercept = -1, linetype = "dashed",
               linewidth = .3) +
    scale_x_continuous(
      breaks = all_per
    ) +
         labs(title = "",
         x = "Event Time",
         y = "Treatment Effect") +
    theme_bw(base_size = 12) +
    theme(plot.title   = element_text(hjust = .25),
          legend.position = "right")
  
  p
```

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

### BibTeX Citation for Research Paper

Source: http://marcelortiz.com/triplediff/authors.html

Use this BibTeX entry to cite the research paper titled 'Better Understanding Triple Differences Estimators'.

```BibTeX
@Article{,
  title = {Better Understanding Triple Differences Estimators},
  author = {Marcelo Ortiz-Villavicencio and Pedro H.C. Sant'Anna},
  year = {2025},
  journal = {ArXiv},
  url = {https://arxiv.org/abs/2505.09942},
}
```

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

### agg_ddd()

Source: http://marcelortiz.com/triplediff/reference/index.html

Aggregate Group-Time Average Treatment Effects in Staggered Triple-Differences Designs.

```APIDOC
## agg_ddd()

### Description
Aggregate Group-Time Average Treatment Effects in Staggered Triple-Differences Designs.
```

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

### Estimate Triple Difference with Covariates

Source: http://marcelortiz.com/triplediff/reference/ddd.html

Demonstrates a two-period DDD estimation using covariates and the DR (doubly robust) method.

```R
#----------------------------------------------------------
# Triple Diff with covariates and 2 time periods
#----------------------------------------------------------
set.seed(1234) # Set seed for reproducibility
# Simulate data for a two-periods DDD setup
df <- gen_dgp_2periods(size = 5000, dgp_type = 1)$data

head(df)
#> Key: <id>
#>       id state partition  time        y        cov1       cov2      cov3
#>    <int> <num>     <num> <int>    <num>       <num>      <num>     <num>
#> 1:     1     0         0     1 209.9152 -0.97080934 -1.1726958 2.3893945
#> 2:     1     0         0     2 417.5260 -0.97080934 -1.1726958 2.3893945
#> 3:     2     0         0     1 211.4919  0.02591115  0.2763066 0.1063123
#> 4:     2     0         0     2 420.3656  0.02591115  0.2763066 0.1063123
#> 5:     3     0         0     1 221.9431  0.97147321 -0.4292088 0.5012794
#> 6:     3     0         0     2 440.9623  0.97147321 -0.4292088 0.5012794
#>          cov4 cluster
#>         <num>   <int>
#> 1:  0.2174955      39
#> 2:  0.2174955      39
#> 3: -0.1922253      29
#> 4: -0.1922253      29
#> 5:  1.1027248      44
#> 6:  1.1027248      44

att_22 <- ddd(yname = "y", tname = "time", idname = "id", gname = "state",
              pname = "partition", xformla = ~cov1 + cov2 + cov3 + cov4,
             data = df, control_group = "nevertreated", est_method = "dr")

summary(att_22)
#>  Call:
#> ddd(yname = "y", tname = "time", idname = "id", gname = "state", 
#>     pname = "partition", xformla = ~cov1 + cov2 + cov3 + cov4, 
#>     data = df, control_group = "nevertreated", est_method = "dr")
#> =========================== DDD Summary ==============================
#>  DR-DDD estimation for the ATT: 
#>      ATT       Std. Error    Pr(>|t|)  [95% Ptwise. Conf. Band]              
#>     -0.0780       0.0828       0.3463      -0.2404       0.0843              
#> 
#>  Note: * indicates that the confidence interval does not contain zero.
#>  --------------------------- Data Info   -----------------------------
#>  Panel Data
#>  Outcome variable: y
#>  Qualification variable: partition
#>  No. of units at each subgroup:
#>    treated-and-eligible: 1232
#>    treated-but-ineligible: 1285
#>    eligible-but-untreated: 1256
#>    untreated-and-ineligible: 1227
#>  --------------------------- Algorithms ------------------------------
#>  Outcome Regression estimated using: OLS
#>  Propensity score estimated using: Maximum Likelihood
#>  --------------------------- Std. Errors  ----------------------------
#>  Level of significance:  0.05
#>  Analytical standard errors.
#>  Type of confidence band:  Pointwise Confidence Interval
#>  =====================================================================
#>  See Ortiz-Villavicencio and Sant'Anna (2025) for details.
```

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

### Estimate ATT(2,2) using ddd

Source: http://marcelortiz.com/triplediff/index.html

Estimates the average treatment effect on the treated using the ddd function with specified formula and control group.

```R
att_22 <- ddd(yname = "y", tname = "time", idname = "id", gname = "state",
             pname = "partition", xformla = ~cov1 + cov2 + cov3 + cov4,
             data = df, control_group = "nevertreated", est_method = "dr")

summary(att_22)
```

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

### Estimate GMM-based DDD with not-yet-treated units

Source: http://marcelortiz.com/triplediff/index.html

Uses the ddd function with control_group set to 'notyettreated' to leverage additional information for estimation.

```R
# do the same but using GMM-based notyettreated
att_gt_nyt <- ddd(yname = "y", tname = "time", idname = "id", gname = "state", 
              pname = "partition", xformla = ~cov1 + cov2 + cov3 + cov4,
              data = data, control_group = "notyettreated", 
              base_period = "universal", est_method = "dr")

summary(att_gt_nyt)
```

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

### ddd()

Source: http://marcelortiz.com/triplediff/reference/ddd.html

Performs Triple Difference estimation on panel data.

```APIDOC
## ddd(yname, tname, idname, gname, pname, xformla, data, control_group, base_period, est_method)

### Description
Estimates the Average Treatment Effect on the Treated (ATT) for groups and time periods using a Triple Difference (DDD) approach.

### Parameters
- **yname** (string) - Name of the outcome variable.
- **tname** (string) - Name of the time variable.
- **idname** (string) - Name of the unit identifier variable.
- **gname** (string) - Name of the group variable.
- **pname** (string) - Name of the partition variable.
- **xformla** (formula) - Formula for covariates.
- **data** (data.frame) - The dataset to be used.
- **control_group** (string) - The control group definition (e.g., "nevertreated").
- **base_period** (string) - The base period strategy (e.g., "varying").
- **est_method** (string) - The estimation method (e.g., "dr").
```

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

### gen_dgp_2periods(size, dgp_type)

Source: http://marcelortiz.com/triplediff/reference/gen_dgp_2periods.html

Generates panel data with a single treatment date and two periods based on the specified size and DGP type.

```APIDOC
## gen_dgp_2periods(size, dgp_type)

### Description
Generates panel data with a single treatment date and two periods. This function is used to create datasets for testing estimators under different nuisance function specifications.

### Arguments
- **size** (Integer) - Number of units.
- **dgp_type** (Integer) - Value in {1, 2, 3, 4}. 1 = both nuisance functions correct; 2 = only the outcome model correct; 3 = only the propensity score correct; 4 = both nuisance functions incorrect.

### Value
A list containing:
- **data** (data.table) - Long format data with columns: id, state, time, partition, x1, x2, x3, x4, y, cluster.
- **att** (Numeric) - True average treatment effect on the treated (ATT), set to 0.
- **att.unf** (Numeric) - Oracle ATT computed under the unfeasible specification.
- **eff** (Numeric) - Theoretical efficiency bound for the estimator.
```

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

### gen_dgp_2periods()

Source: http://marcelortiz.com/triplediff/reference/index.html

Function that generates panel data with single treatment date assignment and two time periods.

```APIDOC
## gen_dgp_2periods()

### Description
Function that generates panel data with single treatment date assignment and two time periods.
```

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

### Aggregate DDD Results

Source: http://marcelortiz.com/triplediff/index.html

Aggregate the group-time average treatment effects into event-study estimates using the agg_ddd function.

```R
es_e <- agg_ddd(att_gt, type = "eventstudy")

summary(es_e)
```

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

### Aggregate DDD results by group

Source: http://marcelortiz.com/triplediff/reference/agg_ddd.html

Performs aggregation of ATT estimates based on cohort or group.

```R
agg_ddd(out, type = "group", alpha = 0.10)
```

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

### Estimate Group-Time Average Treatment Effect

Source: http://marcelortiz.com/triplediff/index.html

Use the ddd function to estimate the ATT(g,t) for panel data. Requires specifying outcome, time, id, group, and partition variables.

```R
att_gt <- ddd(yname = "y", tname = "time", idname = "id", gname = "state", 
              pname = "partition", xformla = ~cov1 + cov2 + cov3 + cov4,
              data = data, control_group = "nevertreated", 
              base_period = "universal", est_method = "dr")

summary(att_gt)
```

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

### agg_ddd(ddd_obj, type, alpha)

Source: http://marcelortiz.com/triplediff/reference/agg_ddd.html

Aggregates DDD results based on the specified type and significance level.

```APIDOC
## agg_ddd(ddd_obj, type, alpha)

### Description
Aggregates DDD results from a previously computed object. Supports event study, group, and calendar time aggregation.

### Parameters
- **ddd_obj** (object) - Required - The DDD object to be aggregated.
- **type** (string) - Required - The type of aggregation to perform: "eventstudy", "group", or "calendar".
- **alpha** (numeric) - Optional - The level of significance for confidence intervals (default is 0.10).
```
Tokens:
Raw