Sample from a multivariate normal distribution to create "in-contact" n point pairs based on real-time-location systems accuracy, and generate a distribution describing observed distances between point pairs.
findDistThresh(
n = 1000,
acc.Dist1 = 0.5,
acc.Dist2 = NULL,
pWithin1 = 90,
pWithin2 = NULL,
spTh = 0.666
)Integer. Number of "in-contact" point-pairs used in the expected-distance distribution(s). Defaults to 1000.
Numerical. Accuracy distance for point 1.
Numerical. Accuracy distance for point 2. If == NULL, defaults to acc.Dist1 value.
Numerical. Percentage of data points within acc.Dist of true locations for point 1.
Numerical. Percentage of data points within acc.Dist of true locations for point 2. If == NULL, defaults to pWithin1 value.
Numerical. Pre-determined distance representing biological threshold for contact.
Output is a list containing 3 named vectors. The first vector describes summary statistics of the simulated distance distribution. The second vector described adjusted spTh values that will capture approximately 84, 98, and 100 spTh value (all calculated using the Empirical rule). Finally, the third vector describes the observed frequency of contact observation given the spTh adjustments listed in the second vector.
This function is for adjusting contact-distance thresholds (spTh) to account for positional accuracy of real-time-location systems, assuming random (non-biased) error in location-fix positions relative to true locations. Essentially this function can be used to determine an adjusted spTh value that likely includes >= 99-percent of true contacts defined using the initial spTh.
NOTE that the description given in our paper, Farthing et al. (2020) (full reference below), for how this function works is slightly incorrect. In the paper, we state that contact definitions should be updated according to an upper 99 distance distribution generated from a multivariate in-contact point distribution. THIS IS NOT CORRECT, however, as this confidence interval relates to determination of the mean of the distance distribution (which is not what we want). Instead, the desired spatial threshold should be updated based on standard deviations away from the mean. Because of the empirical rule, we can capture the spatial threshold values capable of capturing roughly 84, 98 and 100 process of writing another paper addressing this mistake and improving the method.
Farthing, T.S., Dawson, D.E., Sanderson, M.W., and Lanzas, C. 2020. Accounting for space and uncertainty in real-time-location- system-derived contact networks. Ecology and Evolution 10(11):4702-4715.
# NOT RUN {
findDistThresh(n = 10, acc.Dist1 = 0.5, acc.Dist2 = NULL,
pWithin1 = 90, pWithin2 = NULL, spTh = 0.5)
# }
Run the code above in your browser using DataLab