fixes (version 0.4.1)

run_es: Run Event Study with Fixed Effects

Description

Performs an event study analysis using fixed effects regression on panel data. The function automatically generates lead and lag dummies for each relative period around treatment, supports covariates and flexible fixed effects, and allows for clustered standard errors and observation weights.

Usage

run_es(
  data,
  outcome,
  treatment,
  time,
  staggered = FALSE,
  timing,
  lead_range = NULL,
  lag_range = NULL,
  covariates = NULL,
  fe,
  cluster = NULL,
  weights = NULL,
  baseline = -1,
  interval = 1,
  time_transform = FALSE,
  unit = NULL
)

Value

A tibble containing the event study estimates: - `term`: Lead or lag dummy name (e.g., `"lead3"`, `"lag0"`) - `estimate`: Coefficient estimate - `std.error`: Standard error - `statistic`: t-statistic - `p.value`: p-value - `conf_high`: Upper 95

- `conf_low`: Lower 95

- `relative_time`: Relative period (scaled by `interval`) - `is_baseline`: Logical, `TRUE` only for the omitted baseline period

Arguments

data

A data frame containing the panel dataset.

outcome

The outcome variable, unquoted. You can supply a raw variable (e.g., `y`) or a function call (e.g., `log(y)`).

treatment

Treatment assignment indicator (unquoted). Should be binary (`0/1` or logical). Typically equals 1 in and after the treated period, 0 otherwise.

time

The time variable (unquoted). Used for relative period calculation.

staggered

Logical. If `TRUE`, allows each unit to have its own treatment timing (supports staggered adoption). If so, supply `timing` as a variable name. Default is `FALSE`.

timing

If `staggered = FALSE`, supply a single numeric or date value (e.g., `2005` or `"2005-01-01"`). If `staggered = TRUE`, supply the unquoted variable indicating each unit's treatment time. Never-treated units (with `NA`) are allowed and automatically included as controls. If `time_transform = TRUE`, specify `timing` as an integer corresponding to the transformed index.

lead_range

Number of pre-treatment periods (leads) to include (e.g., 5 creates `lead5`, ..., `lead1`). If `NULL`, automatically determined.

lag_range

Number of post-treatment periods (lags) to include (e.g., 3 creates `lag0`, `lag1`, `lag2`, `lag3`). If `NULL`, automatically determined.

covariates

Optional covariates for the regression. Must be a one-sided formula (e.g., `~ x1 + x2`). Default is `NULL`.

fe

Fixed effects specification, as a one-sided formula (e.g., `~ id + year`). Required.

cluster

Cluster-robust standard errors. Accepts a one-sided formula (e.g., `~ id`), or a character vector of column names. Default is `NULL`.

weights

Optional observation weights. Accepts a one-sided formula (e.g., `~ wt`), character string, or bare variable name. Default is `NULL`.

baseline

Which relative period to use as the omitted (reference) period (e.g., `-1`). This dummy is excluded from estimation and added to the results with estimate 0.

interval

The interval between time periods (e.g., `1` for yearly data, `5` for 5-year periods). Default is `1`.

time_transform

Logical. If `TRUE`, time is replaced by a sequential integer per unit (useful for irregular panels). Default is `FALSE`.

unit

Panel unit identifier. Required if `time_transform = TRUE`. Must be an unquoted variable name (e.g., `id`).

Details

This function streamlines event study regression for panel data with flexible support for:

- Staggered adoption: units with `NA` in `timing` are included as controls (all event dummies zero). - Relative time calculation: `(time - timing) / interval`. - Automatic dummy generation for specified leads and lags. - Omission of the baseline period from estimation and its re-addition with estimate 0. - Optional transformation of time to a unit-specific sequence (`time_transform = TRUE`), allowing for irregular or gapped panel structures.

**Collinearity:** If some covariates are perfectly collinear with fixed effects or other regressors, they are automatically dropped from the regression. A message will be displayed listing dropped variables.

Examples

if (FALSE) {
# Minimal use: simple DiD with two-way fixed effects
run_es(
  data = df,
  outcome = y,
  treatment = treat,
  time = year,
  timing = 2000,
  lead_range = 2,
  lag_range = 2,
  fe = ~ id + year,
  baseline = -1
)

# With weights, cluster, and covariates
run_es(
  data = df,
  outcome = y,
  treatment = treat,
  time = year,
  timing = 2000,
  lead_range = 2,
  lag_range = 3,
  covariates = ~ x1 + x2,
  fe = ~ id + year,
  cluster = ~ id,
  weights = ~ popwt,
  baseline = -1
)

# Staggered adoption: timing is unit-specific
run_es(
  data = df,
  outcome = y,
  treatment = treat,
  time = year,
  staggered = TRUE,
  timing = treat_time,
  lead_range = 3,
  lag_range = 4,
  fe = ~ id + year,
  cluster = ~ id,
  baseline = -1
)
}