Spatial Differencing Transformations — spurtransform • spatialunitroot

Applies spatial differencing transformations from Mueller and Watson (2024) to remove spatial unit roots. Supports four transformation types: LBM-GLS, nearest-neighbor, isotropic, and cluster.

Usage

spurtransform(
  formula,
  data,
  coords = NULL,
  prefix = "h_",
  transformation = c("lbmgls", "nn", "iso", "cluster"),
  radius = NULL,
  cluster = NULL,
  latlong = FALSE,
  replace = FALSE,
  separately = FALSE
)

Arguments

formula: A one-sided formula specifying variables to transform, e.g., ~ y + x1 + x2.
data: A data frame containing the variables.
coords: Coordinate specification: a formula (e.g., ~ s_1 + s_2), a matrix, or NULL to auto-detect s_* variables.
prefix: Character prefix for transformed variable names. Default: "h_".
transformation: Type of transformation: "lbmgls" (default), "nn" (nearest-neighbor), "iso" (isotropic), or "cluster".
radius: Radius for isotropic transformation. Required when transformation = "iso".
cluster: Variable name (in data) for cluster transformation. Required when transformation = "cluster".
latlong: Logical; are coordinates latitude/longitude?
replace: Logical; overwrite existing variables with the prefix? Default: FALSE.
separately: Logical; apply transformation separately for each variable (handling missing values per-variable)?

Value

The original data frame with added transformed variables (invisibly).

References

Mueller, U. K. and Watson, M. W. (2024). "Spatial Unit Roots and Spurious Regression." Econometrica, 92, 1661–1695.

Examples

if (FALSE) { # \dontrun{
data(chetty)
transformed <- spurtransform(~ am + fracblack, data = chetty,
                              coords = ~ s_1 + s_2, latlong = TRUE,
                              prefix = "h_")
feols(h_am ~ h_fracblack, data = transformed)
} # }