Abstract
Using estimated parameters log10a and b of Taylor's power law (TPL), different sampling schemes (random, systematic, and clusters), aggregation scales (branch, tree, area, and double area), and sampled sizes (2 to 700) were evaluated in an analysis of an extensive set of counts of the coffee berry borer from the central region of Colombia. Values of b greater than 1 (P < 0.01) were found in more than 90% of the regressions, the values ranging from 0.70 to 2.82 (an average of 1.67), indicating an aggregated pattern of this pest in the field. Values of log10a were from -2.11 to 1.95 (an average of 0.45). When the sample size was small, estimates were corrected for bias. No differences in estimates of either parameter were shown between the cluster and systematic schemes. The random sampling scheme showed marked differences for both log10a and b when compared with the clustering scheme, but there were differences in b but not log10a when compared with the systematic schemes. It was concluded that this was due to the random sampling scheme detecting greater degrees of aggregation, b values being estimated greater than 2.0. Differences in both log10a and b were shown for all changes in scale. Differences in sample size within the different sampling schemes were not reflected in log10a and b values. Possible causes of aggregation of the borer and practical applications of TPL parameters are described.
