An increase in plot size reduces the effect of co-registration errors between field and LiDAR data, reduces the perimeter-to-area ratio, i.e. the edge effect, and reduces sample variance [15], thus increasing model accuracy.
FA models are often been advocated for their parsimony: for problems of relatively high dimensions, reduced sampling variances due to a greatly reduced number of parameters can easily outweigh small biases due to enforcing such structure but, as emphasized above, we need to ensure that the set of factors fitted includes all important factors.
Nonetheless, even if sampling effort is increased to reduce sampling variance, allele frequency changes may be too small to detect in large populations (Hare et al., 2011; Neafsey, 2013).
To reduce sampling variance, the results were obtained by averaging 20 Monte Carlo trials.
Furthermore, possibilities to reduce sampling variance are presented.
To reduce sampling variance, the amino acid probability vector was averaged over the 100 simulation trials.
To reduce sampling variance, an average observed α vector is calculated from 100 simulations.
Focussed COX2 sequencing with larger within-site sample sizes would reduce sampling variance and could provide further insight into the importance of this mutation.
First, α estimates are inherently noisy [ 9] and data on many more genes for each category would be needed to reduce sampling variance around estimates.
Finally, statistical techniques are available - often referred to as regularization methods - which substantially reduce sampling variance, albeit at the expense of introducing some bias, and thus yield 'better' estimates.
