Double sampling revisited.

Abstract

The decision to use double sampling with a regression or ratio estimator is not a simple task. This study was conducted to determine whether a ratio or regression estimator should be used to estimate aboveground biomass of stands dominated by blue grama (Bouteloua gracilis (H.B.K.) Lag ex Steud.) in eastern Colorado. One hundred 0.25-m-1 circular plots were systematically located in a homogeneous stand of blue grama, and on each plot biomass was estimated visually and then clipped. Three methods (classical, jackknife, and bootstrap) of estimating the variance for double sampling with regression and ratio estimator were compared in a simulation study using sample sizes 10, 20, 30, 40, and 50 clipped plots. The ratio estimator consistently had smaller bias and should be used for estimating average clipped weight of blue grama. For n = 10 clipped plots, the jackknife variance estimator is recommended for constructing confidence intervals. For n greater than or equal to 20 clipped plots, the classical variance estimate should be used to obtain reliable estimates of the population variance and in estimating confidence intervals.