EUVIMEDEUVIMED entdecken
Crossref

Two Silent Decisions in Weighting an Opt-In Sample Against a Probability Reference Sample

Juan Pablo Ferreira

Survey Practice · 10.11.2026 · 20 (1) · 1-11

DOI: 10.29115/sp-2026-0052 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Survey Practice, online erschienen am 10.11.2026. Verfasst von Juan Pablo Ferreira. Die Arbeit nennt 1 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Adjusting an opt-in sample against a probability reference sample is now routine, and two decisions inside that adjustment are usually made without being recorded. The first is how the pseudo-weight is formed: whether the reference enters the stacked file with its design weights, and whether the volunteer’s weight is the inverse of the fitted probability or the odds. The second is whether the reference survey’s own sampling variability is allowed into the standard error. Neither is visible afterwards: the achieved margins close to numerical precision either way, and the resulting weight file offers no obvious diagnostic trace of how the pseudo-weight was constructed. Using a simulated opt-in sample adjusted against a stratified probability reference sample, we show that the two wrong constructions of the pseudo-weight fail differently and leave distinct traces in the sum of the weights. The weight-one error is the larger bias in this design and survives a later calibration unless its margins span the reference’s own design strata. Treating the reference’s control totals as fixed divides the estimated standard error by about 1.3 and takes the coverage of a nominal 95% interval from 94 to 78 percent for mass imputation, while leaving the doubly robust interval untouched. And when participation depends on an unobserved variable, every adjustment leaves 4.2 to 5.3 percent of bias, the relative performance changes, with mass imputation becoming the least biased and the doubly robust weight no better than inverse propensity weighting, and no weight-based diagnostic warns of it, while a single held-out benchmark detects the failure in every non-ignorable replication.

Zitieren

APA

Ferreira, J. P. (2026). Two Silent Decisions in Weighting an Opt-In Sample Against a Probability Reference Sample. Survey Practice, 20(1), 1-11. https://doi.org/10.29115/sp-2026-0052

Vancouver

Ferreira JP. Two Silent Decisions in Weighting an Opt-In Sample Against a Probability Reference Sample. Survey Practice. 2026;20(1):1-11. doi:10.29115/sp-2026-0052

BibTeX
@article{Ferreira2026,
  author = {Ferreira, Juan Pablo},
  title = {Two Silent Decisions in Weighting an Opt-In Sample Against a Probability Reference Sample},
  journal = {Survey Practice},
  year = {2026},
  volume = {20},
  number = {1},
  pages = {1--11},
  doi = {10.29115/sp-2026-0052}
}
Kostenlos starten →