Login
Unbiased calibrated estimation on a distribution function in simple random sampling
Résumé The post­stratified estimator sometimes has empty strata. To address this problem, we construct a post­stratified estimator
with post­strata sizes set in the sample. The post-strata sizes are then random in the population. The next step is to construct a smoothed estimator by calculating a moving average of the post­stratified estimators. Using this technique it is possible to 
construct an exact theory of calibration on distribution. The estimator obtained is not only calibrated on distribution, it is linear and completely unbiased. We then compare the calibrated estimator with the regression estimator. Lastly, we propose an approximate variance estimator that we validate using simulations.
   
Mots-clés Unbiased  estimation; Calibration  on  a  distribution  function; Conditional inclusion  probabilities;
Weighting
   
Citation Tillé, Y. (2002). Unbiased calibrated estimation on a distribution function in simple random sampling. Survey Methodology, 28(1), 77-85.
   
Type Article de périodique (Anglais)
Date de publication 24-3-2002
Nom du périodique Survey Methodology
Volume 28
Numéro 1
Pages 77-85
URL http://www.statcan.gc.ca/ads-annonces/12-001-x/6420-eng.pdf