Version 1 (modified by cavallini@…, 10 years ago) (diff)


AniMove Wish List

This is a list of metrics that can be useful. Feel free to comment, and add your own.

Here we try to summarize what is necessary to have a full set of home range analyses in AniMove?. Many of the following analyses are already available, check AnimoveHowto.

Summary Statistics

Number of bearings: Total number of bearings (angles) per dataset

Mean bearing: Mean bearing per dataset (azimuth)

R Concentration of angles: The concentration of angles 1-r is the "circular variance"

Angular deviation: The angular equivalent of linear standard deviation

Rayleigh's z for angles: The z value for Rayleigh's test for significant angles

Duration of Study: Total number of days per dataset

Primary axis angle: The angle which the primary axis is offset from the X axis (90 to -90)

Cramer-von Mises: Test statistic for Complete Spatial Randomness (CSR)(see Cramer-von Mises)

CM heterogeneity p: The probability value for rejecting the null hypothesis of CSR using the Cramer-von Mises

Nearest-Neighbor R: The nearest neighbor test statistic (see Nearest-Neighbor)

Nearest-Neighbor z: The z value of R

Nearest-Neighbor p: The probability value for rejecting the null hypothesis of CSR using Nearest-Neighbor R

Alpha Hull

Code for generating alpha hulls. This example generates hulls for the dataset "Richards Pipit" which consists of two columns, the first being the x coordinate, and the second being the y coordinate for each bird observation. Heuristic rules in this version are 2, 3 and 4 times the average triangle edge length.

Method proposed by Burgman, M.A. & Fox, J.C. (2003). Bias in species range estimates from minimum convex polygons: implications for conservation and options for improved planning. Animal Conservation 6: 19- 28.

k-Nearest Neighbors Convex Hull

A k-NNCH covering for constructing UDs Given a set of specified points the method begins by constructing the convex hull associated with each point and its (k-1) nearest neighbors. We refer to the area covered by the union of all these convex hulls as a k-NNCH covering. We then order the hulls from the smallest to the largest. By progressively taking the union of these from the smallest upwards, until x% of points are included (with some rounding error), we construct the areas whose boundaries represent the x% isolpleth of the densest set of points in our k-NNCH covering. Clément Calenge has written the R code (now included in Adehabitat) to determine the k-nearest-neighbor convex hull following the method by Getz, W.M. & Wilmers, C.C. (2004). A local nearest-neighbor convex-hull construction of home ranges and utilization distributions. Ecography 27: 489-505.PDF.

MCP Sample Size Bootstrap

Bootstraps the points in a dataset given the user- selected parameters to test the effects of sample size on mcp area.

Nearest Neighbor Analysis Test For Complete Spatial Randomness

The Nearest Neighbor Analysis program tests for complete spatial randomness using a selected graphic or polygon feature from a polygon shapefile. Returns a message box displaying the area of the chosen study site and the z and r values.
This program references the FunNNDCSR, which implements the Clark and Evans (1954, Ecology 35. pp445-453) algorithm and allows for either points beyond the boundary to be used for correcting edge effects or uses the correction of Donnely (1978, Holder (eds) Simulation methods in archaeology. pp.91-95). This allows considerable flexibility. If the population has been completely sampled, e.g. animal locations from radio tracking or tag returns, then choose FALSE for correction and ignore the boundary boolean (i.e. set it to TRUE or FALSE as it won't matter). The program defaults to using the boundary correction for anlaysis. If you have not sampled the complete population then select either the edge correction or the boundary correction (if you have sample points beyond the boundary). It checks to see if the sample size is too small (from Donnely 1978) for the normal distribution.

Cramer-Von Mises Test for Complete Spatial Randomness

Description: Works from a menu on either a selected point theme and a selected rectangle graphic or from a rectangle in a shape file. This function is only valid with a rectangular study plot. The wait cursor will appear and then a messagebox will tell you the W value and how many features were accounted for in the analysis. W values relate how clustered or dispersed points are within the graphic rectangle or rectangle theme you specified. A W value of greater than .3 indicates a tendency towards a clumped (clustered) pattern. A W value of 0.3-.06 indicates a random distribution. An R value of less than .06 indicates an organized (uniform) pattern. This statistical test is insensitive to origin but highly sensitive to the upper right corner of the study plot. Its primary usefulness is in detecting heterogeneity that the nearest neighbor analysis (NNA) misses especially NNA base purely on distance measurements. Zimmerman, D.L. 1993, A bivariate Cramer- von Mises type of test for spatial randomness, Appl. Statist.

  1. 42. pp. 43-54.

Circular Point Statistics

Conducts circular point statistics and outputs a graphic representation of the bearings. The red line = mean bearing, which is scaled proportionately to the r value of the angles; r value = 0 to 1, (1=all angles the same) multiplied by the longest line segment.
This implements circular statistics (Batschlet 1981) for the sequence of points in a point coverage. Useful for determining the travel directions from animal movement locations and the significance of direction of travel.. This function requires that the data be ordered in the sequence desired. The function works on the selected records (or all if none are selected) which is useful for examining parts of the movement path. The program will not tell the probability level of rejecting the null hypothesis of directed movement but will give the Z value and sample size which can be looked up in a Z table.

Site Fidelity Test

Creates random angles and uses distances between existing sequential points to determine walk points.
Requires an active point theme. Prompts the user to select one of the following starting points for the simulation: First Point, Last Point, Arithmatic Mean, and Harmonic Mean (Defaults to the First Point).
The simulation compares the observed pattern with a user selectable number of random walks. This uses a Monte Carlo simulation and parameters from the original data to determine if the observed movement pattern has more site fidelity than should occur randomly, is a random pattern or is overly dispersed. It is suggested that a 100 simulations be run first and if the data is close to chosen probability level break point then run it with a 1000 simulations which will more accurately reflect the random walk distribution. Outputs a table with 3 fields, Replicate, R2 and Linearity. Outputs a polyline theme containing polylines of random walks. The polyline attribute table and the R2 table are linked by Replicate and LinkID. If replicates < 100, outputs a chart showing the r2 sorted in ascending order. To identify which r2 belongs with which polyline, select the record from the output table and view the selected polyline in the view. However,as the chart is dynamically linked to the table, selecting individual records will modify the charts color scheme. To save the original chart, add it to a static layout before viewing individual records in the r2 table.The O replicate in the output table and chart is the observed data the others are replicates from simulation runs. Works on selected point records or if none selected the entire table.A single selected polygon or polyline may be use to limit the extent of the random walks. Progress bar tells the progress through the simulation loop.