The function IAB computes the IAB index following the methods described in the paper by
Benhamou et al. (2014). It facilitates global analysis, with the significance testing procedure
described in the paper, but also a local level output, to explore the IAB statistic through time.
IAB(traj1, traj2, tc = 0, dc = 0, local = FALSE, rand = 99)If local=FALSE (the default) IAB returns the numeric value of the IAB index and the associated p-values for one-sided tests for attraction or avoidance.
If local=TRUE IAB returns a dataframe (containing the date/times of all simultaneous fixes (NOTE: times are associated with traj1), along with the distance between fixes at each time , and the IAB index value for each simultaneous fix. A localized p-value (.pa signifies the test for attraction and pb the test for avoidance) and z-score is computed based on rand permutations of the fixes. The pkey columns can be used to match the simultaneous fixes to the original trajectory.
an object of the class ltraj which contains the time-stamped
movement fixes of the first object. Note this object must be a type II
ltraj object. For more information on objects of this type see help(ltraj).
same as traj1.
time threshold for determining simultaneous fixes -- see function: GetSimultaneous.
critical distance where the IAB function will show maximum slope -- see Benhamou et al. (2014) for more advice on selecting this parameter.
logical value indicating whether a dataframe (local = TRUE) containing the IAB
index for each simultaneous fix should be returned (with a local permutation test), or (if local = FALSE - the default)
the global index along with associated global permutation test.
number of permutations to use in the local permutation test.
The function IAB can be used to test for direct interaction in wildlife telemetry data and affords a novel significance testing procedure that takes into account the serially correlated structure of telemetry data. Specifically, it computes an index analogous to the Bhattacharyya coefficient between the potential influence domains of two animals. Like the other indices, IAB is dependent on the selection of an appropriate value for dc (which is termed \(\Delta\) in the article). The dc parameter here is not a threshold distance, but rather the distance at which the function shows maximum slope (see Benhamou et al. 2014).
The significance testing procedure uses a wrapped shifting method in order to maintain the serially correlated structure of the data. At each shift, a sample value of IAB (termed MAB) is computed in order to generate a distribution of values to test against (for more information see Benhamou et al. 2014). Here a local version of this statistical testing procedure is implemented by taking rand samples of the \((n^2 - n\) permutations of unpaired fixes. The p-values are computed following Benhamou et al. (2014), z-scores are calculated based on the mean and standard deviation of this hypothetical distribution.
Benhamou, S., Valeix, M., Chamaille-Jammes, S., Macdonald, D., Loveridge, A.J. (2014) Movement-based analysis of interactions in African lions. Animal Behaviour, 90: 171-180.
GetSimultaneous, DI, Prox
data(deer)
deer37 <- deer[1]
deer38 <- deer[2]
#tc = 7.5 minutes, dc = 50 meters
IAB(deer37, deer38, tc=7.5*60, dc=50)
df <- IAB(deer37, deer38, tc=7.5*60, dc=50, local=TRUE)
Run the code above in your browser using DataLab