Robust / heteroskedasticity-consistent variance for a fitted model

Thin wrapper around sandwich variance estimators. Returns the requested variance-covariance matrix and the corresponding robust standard errors. Supports HC0-HC5 (cross-section), HAC (time-series), and clustered (one-way) variance.

Usage

morie_causal_robust_se(model, type = "HC3", cluster = NULL, ...)

Arguments

model: A fitted model object (typically from stats::lm or stats::glm) compatible with sandwich::vcovHC(), sandwich::vcovHAC(), or sandwich::vcovCL().
type: One of "HC0", "HC1", "HC2", "HC3", "HC4", "HC4m", "HC5", "HAC", or "CL" (clustered). Default "HC3" (Long-Ervin small-sample default).
cluster: Optional one-sided formula or vector identifying the cluster variable (required when type = "CL").
...: Additional arguments forwarded to the chosen sandwich estimator.

Value

Named list with elements vcov (variance matrix), se (named numeric vector of robust SEs), type, and n_coef.

Details

Hard-errors if sandwich is not installed – the HC sandwich algebra is well-tested upstream and re-implementing it inline would be both lengthy and error-prone.

References

Zeileis A, Koll S, Graham N (2020). Various Versatile Variances: An Object-Oriented Implementation of Clustered Covariances in R. Journal of Statistical Software, 95(1), 1-36.

Examples

set.seed(1)
d <- data.frame(x = rnorm(30), y = rnorm(30))
m <- stats::lm(y ~ x, data = d)
n <- 80L
x <- arima.sim(model = list(ar = 0.6), n = n)
y <- as.numeric(x) + rnorm(n, sd = 0.2)
morie_causal_robust_se(m, type = "HC3")
#> $vcov
#>             (Intercept)           x
#> (Intercept)  0.02518345 -0.01119561
#> x           -0.01119561  0.03404757
#> 
#> $se
#> (Intercept)           x 
#>   0.1586929   0.1845198 
#> 
#> $type
#> [1] "HC3"
#> 
#> $n_coef
#> [1] 2
#>