Skip to contents

Coordinated delete-one jackknife across panel waves

Source: R/variance-panel.R
wave_jackknife.Rd

Deterministic counterpart of wave_bootstrap(): builds delete-one-PSU replicate weights that are coordinated across waves (the same PSU is removed from every wave simultaneously), so the sampling covariance the overlap induces is captured without any randomness. Feed the result to change_estimate() for an honest, reproducible variance of a net change. Meant as the cheap exact oracle to validate wave_bootstrap(); it is also a valid variance estimator in its own right.

Usage

wave_jackknife(
  specs,
  strata = NULL,
  psu = NULL,
  refit_steps = "all",
  progress = TRUE
)

Arguments

specs

a named list of weighting_spec objects, one per wave; the names are the wave labels used by change_estimate().

strata, psu

column names of the design strata and PSU, present in every wave's data. The PSU id must be consistent across waves (the same value = the same PSU): that consistency is what makes the coordination possible. strata = NULL treats the whole sample as one stratum; psu = NULL treats each row as its own PSU.

refit_steps

which recipe steps to re-run per replicate; see wave_bootstrap(). "all" (default) re-preps the whole recipe; "calibration" freezes the prefix and re-runs only calibration (StatCan LFS convention).

progress

show a progress message per wave.

Value

an object of class weightflow_wave_jack accepted by change_estimate(); replicate columns are aligned across waves by the union of PSU ids (a PSU absent from a wave, or alone in its stratum, yields that wave's point weights, i.e. a zero jackknife contribution).

Details

Self-contained: each wave's recipe is re-run through prep() on base weights with the deleted PSU zeroed and its stratum mates reweighted by \(n_h/(n_h-1)\); it does not modify bootstrap_weights() or touch R/variance.R.

See also

wave_bootstrap(), change_estimate()

Examples

t1 <- subset(panel_ine, ola == 1 & disp == "R")
t2 <- subset(panel_ine, ola == 2 & disp == "R")
wj <- wave_jackknife(
  list(T1 = weighting_spec(t1, base_weights = w_base),
       T2 = weighting_spec(t2, base_weights = w_base)),
  strata = "estrato", psu = "psu", progress = FALSE)
change_mean(wj, "desocupado")
#> <weightflow net change [coordinated jackknife]>
#>   T1 -> T2
#>   change     : -0.0149512   SE 0.00869933
#>   95% CI    : [-0.0320016, 0.00209913]
#>   V1 0.0001365 | V2 0.0001351 | Cov 9.795e-05 | rho 0.721   (levels)
#>   V = 7.568e-05  vs  V1+V2 = 0.0002716  (deff_change 0.279: overlap saved 72%)