|
|
Sequential sampling is commonly applied within forest inventory, in an effort to minimize
the cost of achieving a nominated standard error of the estimate. Since the stopping
rule is a function of the realized standard error, the sample-based estimates of the
standard error are biased. A simple work-around was suggested by Edelman in 1991,
conditional on the population being normally distributed. I present and discuss the
results of simulation studies that demonstrate that the assumption of normality is key,
and discuss extensions into spatially-correlated populations.
|
|