Estimation with Different Stratification at Two Occasions

Abstract

Abstract Since forest surveys may use different strata over time, permanent plots may be remeasured and new plots may be selected with different probabilities. The inclusion probabilities and joint probabilities of selecting both remeasured and new plots, considering both occasions, are derived. The Horvitz-Thompson estimator using two approximations to actual probabilities of selection and a stratified sampling estimator ignoring initial probabilities of selection are compared for bias and efficiency. The Yates-Grundy, probability proportional to size with replacement, jackknife, and stratified sampling variance estimators are evaluated in terms of bias in estimating the true variance of the estimators. Three populations are constructed with different rates of movement between strata from time 1 to time 2 to evaluate the estimators. The stratified sampling estimator performed poorly. The jackknife variance estimator is recommended. FOR. SCI. 39(2):368-382.