USGS Home
Contact USGS
Search USGS

On the basis of our evaluations, if the study area habitat is heterogeneous and reliable current information is available, managers should stratify the study area and use an approximate optimal allocation. However, managers should be alert to problems with confidence interval coverage. Simple random sampling without replacement with the simple estimator is preferable unless the correlation between area and count is > 50% the CV of the area divided by the CV of the count (Cochran 1977:158). Under these conditions, the ratio estimator has smaller variance. Cochran (1977:165) stated that the separate ratio estimator was essentially unbiased when sample sizes were large enough in each strata for the variance formula to be valid for each stratum and when the square root of the number of strata times the CV of the mean area did not exceed 0.3. With any sample size, the pps estimator with PPS sampling is unbiased. With small sample sizes and greater sampling intensity, the combined ratio estimator may give precise estimates. Variables other than area, such as the amount of preferred habitat on each sampling unit or the number of animals present during a previous survey, should also be examined to see how they correlate with the count on the sampling unit. If animals tend to concentrate in 1 stratum, then complete counting in that stratum may give precise estimates but poor confidence interval coverage for some combinations of sampling plans and estimators.

Choosing a sampling plan and estimator often has been based on familiarity rather than theory. Although no single combination of sampling plan and estimator is best for all situations, an informed choice can be made. Once the relative importance of accuracy, confidence interval coverage, and cost is determined, then results of our study can be used to help managers decide which sampling plan and estimator is most appropriate. Nonetheless, managers should be cautious in using any sampling plan on aggregated populations.