### Install bacondecomp

Source: https://github.com/evanjflack/bacondecomp/blob/master/README.md

Installation instructions for the package from CRAN or GitHub.

```R
install.packages("bacondecomp")
```

```R
library(devtools)
install_github("evanjflack/bacondecomp")
```

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

### Install bacondecomp Development Version from GitHub

Source: https://context7.com/evanjflack/bacondecomp/llms.txt

Install the latest development version of the bacondecomp package directly from GitHub using the devtools package. This is useful for accessing the newest features or bug fixes.

```r
library(devtools)
install_github("evanjflack/bacondecomp")
```

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

### Install bacondecomp from CRAN

Source: https://context7.com/evanjflack/bacondecomp/llms.txt

Use this command to install the stable version of the bacondecomp package from the Comprehensive R Archive Network (CRAN).

```r
install.packages("bacondecomp")
```

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

### Perform Controlled Goodman-Bacon Decomposition

Source: https://context7.com/evanjflack/bacondecomp/llms.txt

This example shows how to perform a Goodman-Bacon decomposition that includes time-varying covariates. The `bacon()` function returns a list containing the within-group coefficient, its weight, and the between-group 2x2 estimates. This allows for a more nuanced analysis by accounting for additional sources of variation.

```r
library(bacondecomp)

# Decomposition with time-varying controls
# Including log population and log income as controls
ret_bacon <- bacon(
  formula = l_homicide ~ post + l_pop + l_income,
  data = bacondecomp::castle,
  id_var = "state",
  time_var = "year"
)

# Access components of the controlled decomposition
beta_hat_w <- ret_bacon$beta_hat_w    # Within timing group coefficient
Omega <- ret_bacon$Omega              # Weight on within-group estimate
two_by_twos <- ret_bacon$two_by_twos  # Between-group 2x2 estimates

print(paste("Within timing group estimate (beta_hat_w):", round(beta_hat_w, 4)))
print(paste("Within group weight (Omega):", round(Omega, 4)))

# Calculate weighted sum using the formula:
# beta_hat = Omega * beta_hat_w + (1 - Omega) * sum(s_kl * beta_hat_d_bkl)
beta_hat_b <- weighted.mean(two_by_twos$estimate, two_by_twos$weight)
bacon_coef_cont <- Omega * beta_hat_w + (1 - Omega) * beta_hat_b
print(paste("Weighted sum of decomposition =", round(bacon_coef_cont, 4)))

# Verify with two-way FE including controls
two_way_fe_cont <- lm(l_homicide ~ post + l_pop + l_income +
                        factor(state) + factor(year),
                      data = bacondecomp::castle)
print(paste("Two-way FE estimate =",
            round(two_way_fe_cont$coefficients["post"], 4)))

```

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

### Summarize Bacon Decomposition Results

Source: https://context7.com/evanjflack/bacondecomp/llms.txt

Aggregates 2x2 decomposition results by comparison type to show total weight and average estimate. Use `return_df = TRUE` to get results as a data frame for further analysis.

```r
library(bacondecomp)

# Perform decomposition
df_bacon <- bacon(
  formula = l_homicide ~ post,
  data = bacondecomp::castle,
  id_var = "state",
  time_var = "year",
  quietly = TRUE
)

# Get summary by comparison type
bacon_summary(df_bacon)
#>                       type  weight  avg_est
#> 1 Earlier vs Later Treated 0.05041 -0.00143
#> 2 Later vs Earlier Treated 0.05041  0.11706
#> 3     Treated vs Untreated 0.89918  0.07965

# Return as data frame for further analysis
summary_df <- bacon_summary(df_bacon, return_df = TRUE)

# Identify problematic comparisons
# "Later vs Earlier Treated" can be biased if treatment effects vary over time
print(paste("Weight on potentially biased comparisons:",
            round(summary_df[summary_df$type == "Later vs Earlier Treated", "weight"], 4)))

```

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

### bacon_summary()

Source: https://context7.com/evanjflack/bacondecomp/llms.txt

Aggregates 2x2 decomposition results by comparison type to calculate total weights and weighted average estimates.

```APIDOC
## bacon_summary(df_bacon, return_df = FALSE)

### Description
Aggregates the 2x2 decomposition results by comparison type, providing the total weight and weighted average estimate for each category.

### Parameters
#### Request Body
- **df_bacon** (object) - Required - The output object from the bacon() function.
- **return_df** (boolean) - Optional - If TRUE, returns the summary as a data frame instead of printing to console.

### Response
#### Success Response (200)
- **type** (string) - The comparison category (e.g., Earlier vs Later Treated).
- **weight** (numeric) - The total weight assigned to this comparison type.
- **avg_est** (numeric) - The weighted average estimate for this comparison type.
```

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

### Plot 2x2 Estimates and Weights

Source: https://github.com/evanjflack/bacondecomp/blob/master/index.ipynb

Visualize the decomposition by plotting each 2x2 estimate against its corresponding weight. This helps identify which specific comparisons drive the overall fixed effects estimate.

```R
ggplot(df_bacon) +
  aes(x = weight, y = estimate, shape = factor(type)) +
  labs(x = "Weight", y = "Estimate", shape = "Type") +
  geom_point()
```

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

### Perform Basic Goodman-Bacon Decomposition

Source: https://context7.com/evanjflack/bacondecomp/llms.txt

This snippet demonstrates how to perform a basic Goodman-Bacon decomposition using the `bacon()` function. It estimates the effect of 'post' on 'l_homicide' using the 'castle' dataset, identifying units by 'state' and time by 'year'. The output shows 2x2 estimates and their weights.

```r
library(bacondecomp)

# Basic decomposition without controls (Castle Doctrine data)
# Estimate effect of castle doctrine laws on homicide rates
df_bacon <- bacon(
  formula = l_homicide ~ post,
  data = bacondecomp::castle,
  id_var = "state",
  time_var = "year",
  quietly = FALSE
)

# View the 2x2 estimates and weights
head(df_bacon)

```

```r
# Verify decomposition equals TWFE estimate
coef_bacon <- sum(df_bacon$estimate * df_bacon$weight)
print(paste("Weighted sum of decomposition =", round(coef_bacon, 4)))

# Compare with standard two-way fixed effects
fit_tw <- fixest::feols(l_homicide ~ post | state + year,
                        data = bacondecomp::castle)
print(paste("Two-way FE estimate =", round(fit_tw$coefficients[1], 4)))

```

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

### Built-in Datasets

Source: https://context7.com/evanjflack/bacondecomp/llms.txt

Access to panel datasets provided by the package for replication and analysis.

```APIDOC
## Built-in Datasets

### Description
The package includes several datasets used in seminal difference-in-differences research.

### Available Datasets
- **castle** (data.frame) - State-year panel data from Cheng and Hoekstra (2013) on castle doctrine laws.
- **math_reform** (data.frame) - Aggregated state-cohort data from Goodman (2019) on education reform.
- **divorce** (data.frame) - State-year panel data from Stevenson and Wolfers (2006) on unilateral divorce laws.

### Usage Example
```r
data(castle)
head(castle)
```
```

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

### Visualize Decomposition Results

Source: https://github.com/evanjflack/bacondecomp/blob/master/README.md

Create a scatter plot of weights versus estimates using ggplot2.

```R
library(ggplot2)

ggplot(df_bacon) +
  aes(x = weight, y = estimate, shape = factor(type)) +
  geom_point() +
  geom_hline(yintercept = 0) + 
  theme_minimal() +
  labs(x = "Weight", y = "Estimate", shape = "Type")
```

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

### Load and Inspect Castle Doctrine Data

Source: https://context7.com/evanjflack/bacondecomp/llms.txt

Loads the 'castle' dataset, which contains state-year panel data on the effect of 'castle doctrine' laws on homicide rates. Displays the first few rows with relevant columns.

```r
library(bacondecomp)

# Load castle doctrine data
data(castle)
head(castle[, c("state", "year", "l_homicide", "post", "l_pop", "l_income")])
#>   state year l_homicide post    l_pop l_income
#> 1    AK 2000   2.227751    0 13.47196 10.44691
#> 2    AK 2001   1.945910    0 13.47946 10.47745
#> 3    AK 2002   2.063070    0 13.48688 10.44863
#> 4    AK 2003   2.010792    0 13.49246 10.47087
#> 5    AK 2004   1.791759    0 13.49907 10.51173
#> 6    AK 2005   1.609438    0 13.50535 10.56039

# Basic analysis
df_bacon <- bacon(l_homicide ~ post, data = castle,
                  id_var = "state", time_var = "year")

```

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

### Load and Analyze Math Reform Data

Source: https://context7.com/evanjflack/bacondecomp/llms.txt

Loads the 'math_reform' dataset, which contains aggregated state-cohort data on the effect of compulsory high school math coursework on future earnings. Performs a bacon decomposition and displays a summary.

```r
library(bacondecomp)

# Load math reform data
data(math_reform)
head(math_reform)
#>   state class reform_math reformyr_math incearn_ln
#> 1    AK  1973           0            NA   10.19299
#> 2    AK  1974           0            NA   10.19891
#> 3    AK  1975           0            NA   10.18728

# Decomposition for education reform effect
df_bacon <- bacon(
  formula = incearn_ln ~ reform_math,
  data = math_reform,
  id_var = "state",
  time_var = "class"
)

# View summary
bacon_summary(df_bacon)
#>                       type  weight  avg_est
#> 1 Earlier vs Later Treated 0.06353  0.02868
#> 2 Later vs Earlier Treated 0.05265  0.03375
#> 3     Treated vs Untreated 0.88382 -0.00129

```

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

### Load Divorce Law Data

Source: https://context7.com/evanjflack/bacondecomp/llms.txt

Loads the 'divorce' dataset, which contains state-year panel data examining the effect of unilateral divorce laws. The structure of the data is then displayed.

```r
library(bacondecomp)

# Load divorce data
data(divorce)
str(divorce)

```

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

### Visualize Bacon Decomposition Estimates vs Weights

Source: https://context7.com/evanjflack/bacondecomp/llms.txt

Plots 2x2 decomposition estimates against their weights, colored by comparison type. This helps identify which comparisons drive the overall TWFE estimate and detect concerning patterns. Use `ret_bacon$two_by_twos` for controlled decomposition plots.

```r
library(bacondecomp)
library(ggplot2)

# Perform decomposition
df_bacon <- bacon(
  formula = l_homicide ~ post,
  data = bacondecomp::castle,
  id_var = "state",
  time_var = "year"
)

# Plot estimates vs weights by type
ggplot(df_bacon) + 
  aes(x = weight, y = estimate, shape = factor(type), color = factor(type)) + 
  geom_point(size = 3) + 
  geom_hline(yintercept = 0, linetype = "dashed", color = "gray50") + 
  theme_minimal() + 
  labs(
    x = "Weight",
    y = "Estimate",
    shape = "Comparison Type",
    color = "Comparison Type",
    title = "Goodman-Bacon Decomposition",
    subtitle = "Castle Doctrine Effect on Homicide Rates"
  ) + 
  theme(legend.position = "bottom")

# For controlled decomposition
ret_bacon <- bacon(
  formula = l_homicide ~ post + l_pop + l_income,
  data = bacondecomp::castle,
  id_var = "state",
  time_var = "year"
)

ggplot(ret_bacon$two_by_twos) + 
  aes(x = weight, y = estimate, shape = factor(type)) + 
  geom_point(size = 3) + 
  geom_hline(yintercept = 0, linetype = "dashed") + 
  theme_minimal() + 
  labs(x = "Weight", y = "Estimate", shape = "Type")

```

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

### Calculate Two-Way Fixed Effects Estimate

Source: https://github.com/evanjflack/bacondecomp/blob/master/index.ipynb

Fit a standard two-way fixed effects model using `lm` to obtain the overall estimate. This estimate should be equal to the weighted average of the 2x2 decompositions.

```R
fit_tw <- lm(l_homicide ~ post + factor(state) + factor(year), 
             data = bacon::castle)
print(paste("Two-way FE estimate =", fit_tw$coefficients[2]))
```

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

### Identify Estimate with Top Weight

Source: https://github.com/evanjflack/bacondecomp/blob/master/index.ipynb

Extract and print the specific 2x2 estimate that has the largest weight in the decomposition. This can highlight the most influential comparison in the analysis.

```R
top_weight <- df_bacon[order(-df_bacon$weight), ][1, ]
print(top_weight)
```

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

### Decompose Two-Way Fixed Effects

Source: https://github.com/evanjflack/bacondecomp/blob/master/README.md

Calculate 2x2 differences-in-differences estimates and generate a summary of the decomposition.

```R
library(bacondecomp)
#> Loading required package: fixest
#> fixest 0.9.0, BREAKING changes! (Permanently remove this message with fixest_startup_msg(FALSE).) 
#> - In i():
#>     + the first two arguments have been swapped! Now it's i(factor_var, continuous_var) for interactions. 
#>     + argument 'drop' has been removed (put everything in 'ref' now).
#> - In feglm(): 
#>     + the default family becomes 'gaussian' to be in line with glm(). Hence, for Poisson estimations, please use fepois() instead.
df_bacon <- bacon(incearn_ln ~ reform_math,
                  data = bacondecomp::math_reform,
                  id_var = "state",
                  time_var = "class")
                  
# All 2x2 Comparisons
head(df_bacon)
#>    treated untreated    estimate      weight                     type
#> 2     1987      1985  0.04575585 0.031655309 Later vs Earlier Treated
#> 4     1986      1985  0.11390411 0.001978457 Later vs Earlier Treated
#> 5     1984      1985  0.12012908 0.001978457 Earlier vs Later Treated
#> 6     1985      1987  0.01021379 0.039569136 Earlier vs Later Treated
#> 9     1986      1987 -0.09374495 0.005440756 Earlier vs Later Treated
#> 10    1984      1987  0.09784857 0.013354583 Earlier vs Later Treated
# Summary of Early vs. Later, Later vs. Earlier, and Treated vs. Untreated
bacon_summary(df_bacon)
#>                       type  weight  avg_est
#> 1 Earlier vs Later Treated 0.06353  0.02868
#> 2 Later vs Earlier Treated 0.05265  0.03375
#> 3     Treated vs Untreated 0.88382 -0.00129
```

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

### Decompose Two-Way Fixed Effects Model with bacon

Source: https://github.com/evanjflack/bacondecomp/blob/master/index.ipynb

Use the `bacon` function to decompose a two-way fixed effects model. Specify the outcome and treatment variables, the data, and the id and time variables. This function calculates all 2x2 estimates and their weights.

```R
library(bacon)
library(ggplot2)
df_bacon <- bacon(l_homicide ~ post,
                  data = bacon::castle,
                  id_var = "state",
                  time_var = "year")
```

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