A simulation study assessing the effect of sampling for predictor variable values on estimates of yield

Publisher Website
Google Scholar
Abstract
In a study aimed at investigating the effects of sampling for predictor variable values on yield estimates, independent simple random and stratified random samples of stand density and tree height were drawn from simulated pine plantations containing nine strata. Sample outcomes were substituted into a selected yield function to produce estimates of mean yield. Stratifying by both site index and numbers of trees per unit area improved the precision of the yield estimates by 2/3 over simple random samples. Total sample size had no effect on the magnitude of the improvements in the yield estimate. Site index was the more important of the two stratifying variables, but synergistic effects between the two characteristics for stratification were found. Estimating the mean yield with the average of the sample plot yields is recommended over using the alternative of using the mean values of the independent variables. Averaging the sample plot yields produced superior mean yield estimates with less bias and with smaller average error sizes especially when simple random samples were drawn.

This publication has 0 references indexed in Scilit: