TABLE 1

Accuracy and precision of estimates of the population variance for different sampling designs

NSystematicRandomClustered
Estimated population variance 101.0191.0260.584
 201.0031.0130.757
 501.0111.0330.904
1001.0070.9970.969
Standard deviation of estimates 100.4770.4780.810
 200.2880.3300.616
 500.1510.1920.401
1000.0240.1310.284
Proportion of autocorrelated pairs 100.000.0160.448
 200.000.0160.218
 500.000.0160.089
1000.000.0160.048
  • The autocorrelated data were generated by dividing an ordered vector of 500 random values from a standard normal distribution into groups of five consecutive values. The entire groups and the values within each group were reordered at random, representing a hypothetical transect where always five neighboring locations would show very similar values, with random steps between groups. The data set has an expected variance of 1.0 and the proportion of (autocorrelated) within-group comparisons is 0.016. Three types of samples were taken from the transect: (i) a random sample from all locations, (ii) a systematic sample selecting every fifth location, and (iii) a clustered sample, where entire groups of five neighboring locations were selected at random. Data simulation and sampling were repeated 1000 times for each sample size of 10, 20, 50, or 100.