Fractal.
A program to analyze movement paths of animal movement paths.
by Vilis O. Nams. vnams@nsac.ca
This is a windows-based program that also plots results and the original movement path, saves results to a file, prints plots, etc (all of the pretty things you expect from windows programs).
Fract.zip- - size 474 KBytes, version 5.18, last revised Feb 25, 2009
This is meant as a research tool and has been tested thoroughly. However, I cannot guarantee that all bugs have been found - if you find any then email vnams@nsac.ca about them and I will fix them.
** you cannot use Lat-Long coordinates in any movement path analysis because they are not square coordinates (unless you work on the equator).**
Fractal does the following:
1. Estimates fractal dimension (D) using the basic divider method (as introduced by Mandelbrot).
2. Estimates fractal D, using a resampling version of the divider method, which also estimates fractal D as a function of scale. (Nams 2006: Acta Biotheoretica 54:1-11)
3. Calculates the VFractal estimator, which estimates fractal D as a function of scale but which also provides error estimates (Nams, 1996: Landscape Ecology 11:289-297).
4. Uses the VFractal to calculate other estimators that are useful in analyzing how animals react to spatial scale (Nams 2006: Animal Behaviour. 72: 1197-1203).
5. Estimates Net distance2 travelled as a function of number of steps, and uses this for a statistical test for deviations from a CRW model (Nams and Bourgeois 2004: Can J. Zool, 82:1738-1747).
6. Carries out the scaling test for oriented movement - this is used to test whether a movement path is oriented (Nams 2006: Animal Behaviour. 72: 1197-1203).
7. Transforms movement paths by discretizing or rotating them.
** note you can get these papers from here.
Fractal is distributed for free, with the following conditions:
1. That if you use it, you mail me postcard with a real stamp (i.e. not a meter stamp) on it. This is for my wife, who collects stamps - she needs something for the time I spend on projects like this.
2. That you cite it if you publish research that uses it.
3. That you send me reports of any problems with it (I revise this program often, depending on the current state of my research program, and comments from users).
4. That you send me citations of any papers you publish that use the program.
5. That you do not post it on any web sites - instead, add links to this site. This will ensure that only the latest version is posted.
How to install:
1. Download the file Fract.zip into the directory you want it to reside in.
2. Decompress it using an unzip program.
3. You don't need any installation program. Just run the Fractal.exe file directly.
Fract.zip contains 4 files:
· Fractal.exe - the fractal program.
· Fractal.hlp, Fractal.cnt - the help files
· readme - a short file containing this information that you are reading.
Versions:
vers 5.18 Feb 25/2009
- added the ability to turn off the "Path too short" message
vers 5.19 Feb 25/2009
- inserted tab after = sign in some of output: (mean FractalD, mean FractalD
tranformed, median Fractal D)
vers 5.17 Feb 24/2009
- how to tell when path ends: now three possibilites:
1) in first line of data, anywhere after the x,y locations
2) in a line between the data sets
3) in last line of data, anywhere after the x,y locations
- when combining data, add Median Fractal D and transformed (which is Log(D),
and then transformed back). This is useful when have many short paths, which may
vary greatly in fractal D
vers 5.16 Feb 20/2009
- fixed bugs: wasn't adding label when had many paths in many files, and
combining results
-in above situation also was not combining the paths after first file
vers 5.15 Sept 25/2008
- for Oriented analysis: CRWDiff statistic was wrong. In denominator, had used
incorrect value for maxR2 (Nams 2006; equation 3).
- also added plot of RWDiff. This is also used in the NetDistance estimator (Nams and Bourgeois 2004). Both RDiff and CRWDiff are zero when path is a CRW,
and +ve when oriented. But they differ in precision in various circumstances.
Thus if either is not zero, then the path is not a CRW
vers 5.14 Sept 19/2008
- merged versions 5.13 and 5.13b
- bug: filename for file containing list of files was not saved when clicking
an item in Options/Load
vers 5.13b Sept 15/2008
- bug fix: In some cases wasn't saving to individual files.
vers 5.13 July 23/2008
- added ability to select only those paths within set limits for min and max
length
- bug fix: when combining results, if not possible to estimate overall
FractalD, then a D-value of 1 was still used in the combining
- bug fix: when combining results and weighting means by N, the variable Npaths
was also weighted by N
vers 5.12 June 25/2008
- bug fix: when combining results of >1 path, and saving results to a file,
then Fractal could not read in the results if 0<N<1. Set N=1 in this
situation.
vers 5.11 Oct 11/2007
- with some combinations of settings for analysing many paths, wasn't saving
individual results
- when there were very very small step lengths as compared to the sizes of the
coordinates, rounding error caused Fractal to get stuck in an infinite loop
- couldn't halt program when got stuck in above situation
vers 5.10 July 9/2007
- fixed bug in CRWDiff statistic. Wording in equation in paper (Nams 2006
Animal Behaviour; equation 4) was wrong.
- in denominator was n2l2 - the l should have been mean
step length at the scale of n, not just overall mean step length.
- thus this also affects the Scaling test. But note that this does not
affect the point at which the CRWDiff statistic crosses the axis, which is the
key part of the Scaling test
vers 5.09 April 13/2007
- was not showing discretized expected random for net distance
vers 5.08 March 3/2007
- only show "Use Time" checkbox if are running Vfractal
- also, only read in time if are running VFractal, and if checked UseTime
- did this by creating a function UseTime, and changing variable UseTime to
UseTimeV
- also needed to add ShowObjects to TLoadingOptions on show method (had been
commented out)
vers 5.07 Feb 22/2007
- fixed bug - was showing plots for each path when combining paths when one
path per file
vers 5.06 Dec 19/2006
- fixed bug - sometimes name of path lost when changing Options/Load values
- fixed bug - bad fonts for results screen
vers 5.05 Nov 28/2006
- better error messages when analysis doesn't work
- better error handling with too-short paths
- fixed bug - in combined analysis with Oriented and NetDistance, had not set
AutoMin, AutMax scales to False
- fixed bug - for NetDist and Directed, in Prob of Same direction, better
handling of missing values
- added ability to change fonts for results page
vers 5.04 Nov 15/2006
- when time read in, added various speed and time summary stats
vers 5.03 Nov 8/2006
- added ability for text to separate paths to be at end of data line
vers 5.02 July/2006
- added discretized stat for directed walk
- added ability to set whether are starting each V of VFractal on a step
- bug in Transform Paths had given unequal step lengths for distance
transformation
- fixed display of Display Options - Transform Path page wasn't showing
vers 5.00 Feb/2006
- added scaling test for oriented movement
- added transform paths
vers 4.11 Dec/2005:
- when many paths many files, have pausing to stop between files, not paths
vers 4.10 Aug 23/2005:
- bug fix: when batch name file in different directory than previously-accessed
data file, then first output file had been placed in wrong directory
- updated help file
vers 4.09 Nov 4/2004:
- fixed a bug in displaying basic Fractal estimator - screen updated too
often, slowing down program
vers 4.08 June 30/2004
- fixed a bug: in NetDistance, when discretizing by changing to a fixed step
length, had problems when desired length was smaller than one of the steps
vers 4.07 April 7/2004
- fixed a bug: when calculating min scale size, when some steps are zero
length, sometimes gave error
vers 4.06 Feb 16/2004
- fixed a bug that had displayed too many messages when analyzing many short
paths
vers 4.05
- changed x-axis title for NetDistance estimators
vers 4.04 Jan 22/2004:
- allow drawing movement paths with many data points. Do this by skipping data
points if have more than 10,000.
- introduce calculation of time vs spatial scale. Either assume same time per
step, or time data are in form: x,y,time
vers 4.03 Jan 20/2004:
- fixed a bug introduced in 4.02 that had slowed down the program
vers 4.02 Jan 15/2004:
- removed path length limitations
- added more run-time messages to allow user to know if Fractal is still
running
- increased Max # of spatial scales from 90 to 200
vers 4.01 Jan 8/2004:
- bug fix: couldn't read in text files when lines separated by line-feed only
vers 4.00 April 28/2003:
- added ability to combine results when analysing several movement paths
- misc changes to menus
- changed FractalMean algorithm to only traverse the path twice, once from each
end
vers 3.24 Feb 14/2003:
- fixed two bugs in the fractal routines. Tested effects of these two bugs by
running several analyses on paths, using new vs old version: fixed version gave
D's slightly larger (~0.01-0.02 more), in a haphazard fashion. Thus, bug had
not created significant problems.
- for Net Distance, added the ability change the step size of the movement path
before analysis
- change format of fractal.ini file to Windos format
- FINALLY found the source of the occassional Allocation Error when closing
plot windows
vers 3.23 Jan 7/2003:
- bug in batch run run-no labeling: had used name of file out, rather than
file in
vers 3.22 Nov 22/2002:
- in batch mode can add label for data for each run
vers 3.21 Oct 10/2002:
- data values can now be separated by commas also
vers 3.20 March 25/2002:
- changed message when movement path file is empty - instead of previous
cryptic error message
vers 3.19 Feb 9/2001
- fixed bug in NetDistance, incorrect N for each size
- in output, added: Min,Max Scale, and N Scales that was set up, and added
Random Seed
- fixed bug where Fractal stopped sometimes when a plot was removed
- # of steps - was one too many. Had been counting # of data points
vers 3.18 Oct 4/2000
- added ability to set random number seed
- added expected random walk
- added pause after each batch
- added plot n for netdistance
- added ci and randomwalk to NetDist2 plots
- fixed bug: removed min and max check for MinScale and MaxScale in
Options/Setup. previously had not allowed MinScale to be <1. Don't know why
- fixed bug: couldn't switch off NetDist2 plotting
vers 3.17 July 20/2000
- fixed bug: error check for ranges of min/max for scaling shouldn't have been
run for auto
3.16 - 18-May-2000: bug fixes:
- didn't read in last data line if no blank line afterwards
- for NetDistance estimator, had calculated auto min&max scales based on net
distance, rather than # of steps. Now, min is 1 and max is # steps
- in auto scaling, when removed duplicate step sizes, had dropped all the ones
of that size. This would only have been seen in NetDistance, since other step
sizes are not integer
- in SegmentedCorr, had previously removed CVD estimates from output. Now
removed from plotting and help file also
3.15 - 27-Jan-2000
- bug fix in FractalMean. Window ranges had been saved and displayed
incorrectly, but were used in calculations correctly. Help file was also
incorrect (will be changed soon). Correct ranges are SizeMin = size/(1-window),
and SizeMax = size*(1+window).
3.14 - 20-July-99
- bug fixed which had been created in the previous version. Reading in of some
results files had caused an error
3.13 - 11-May-99
- bugs fixes: first x,y coordinate had been dropped if was 0,0
- when measuring path length, first step had been dropped
- output for saving and displaying results was mixed for last line when many
digits
3.12 - 12-Feb-99
- ability to handle no travel between coordinates
- fixed bug: when reading results from FractalMean, gross distance and not been
converted from log scaling.
- added ability to have tabs between x,y values read in }
3.11 more digits for output results for sizes and N
3.10 - 27-Nov-98
- add ability to exit from options screens by using Esc key
- fixed bug which caused crashes with x,y coordinates that contained many
(10-11) digits
3.09 - 24-Jul-98
- for batch mode, added ability to put all movement paths in one file
- b/c of this, split Options/File into Options/Save and Options/Load.
3.08 - 24-Jul-98
- moved manual from separate, Word file, into help system
3.07- 9-March-98
- removed bug in batch mode, where incorrect directory was sometimes searched
3.06 - 24-Sep-97
- added batch mode capability
3.05- Aug 25/97
- changed VFractal estimator - now each V starts on a movement path step.
Without this, corr cos was slightly positively biased at small scales.
- changed segmented corr cos (in VFractal) - had a slight +ve bias at small #
Max segment lengths. Used a different estimate for mean in that calculation.
Derivation for this will be given in Nams 1998.
- added a few more summary statistics to main results
- only do 32-bit compilation of this version
- moved Scaling options to a separate screen - Options screen was getting too
crowded
3.04- May 5/97
- added ability to place movement plot in main screen
- added probability of same turns
- changed titles of some of the plot headings ("VFractal - NetCos"
->"VFractal - Fractal Dimension"
3.03- Feb 25/97 - bug fix: Fractal had displayed Log(grossdist) instead of Distance in plots
3.02- Feb 11/97 - fixed bug that caused math error when reading in past results file
3.01- Jan 24/97
- added segmented correlation
- changed name: Segmented Fractal had been called FractalBootstrap. Was not
intuitive name.
- name of data file to open had not saved. Fixed it.
3.0 - Windows version.
- Oct 5/96 - version 3.0 first released
2.0 - DOS user interface added for external use.
1.0 - Original DOS version, for internal use.