Changes in DENSITY 3.3
Version 3.3 is mostly a consolidation of Version 3.2. Some bugs are fixed and some minor new features added.
New detection function
30 03 05 Negative exponential detection function added as an option
(Options | Inverse prediction)
02 04 05 New output field 'DetectFn' to record the detection function used
in analysis (i.e. moved from header to each output record)
Target values
Changes to this input format in Version 3.2 were inadequately tested.
A bug caused a spurious error if Options were changed after 'Read data',
and the screen and log output included empty fields - a hangover from
the previous format. These errors were corrected 28 02 05.
Initial values
01 03 05
The 'initial values' panel has been hidden on the main screen. It is replaced
by a read-only panel. Initial values are now set exclusively on the Inverse
Prediction page in Options. The default setting has been changed to 'Auto'.
Coding has been streamlined, so some bugs may yet appear.
Power analysis fields AutoD, Autog0, AutoSig now record the actual values used for IP or ML at each replicate, rather than values from a separate call to the algorithm.
New simulation output field
06 04 05 Added simulation field CumTS% for cumulative percentage
of traps or bait stations visited over a trapping session
Preview of simulation settings
18 03 05 Tired of running simulations in PowerAnalysis only to discover
the next day that you had forgotten a critical setting or output field?
Help is at hand. Before you click 'GO', use Tools | Preview to see most
of the header information that will appear in your output file. Seeing
your settings in this form may alert you to omissions or mistakes.
Maximum likelihood estimation of density
18 02 05 2-D Integration Region (buffer beyond traps)
'Fixed' option may specify multiple levels in metres. Separate levels with
commas and/or whitespace:
e.g. 100, 150, 200, 250
'Multiple of initial sigma' option may specify levels in units of the initial
value of sigma.
e.g. 3,5,7
Analyses are repeated for each level of buffer width ('MLBuff' in the output).
This provides a valuable check on the algorithm. Ideally the fitted parameters
are independent of the chosen buffer width, at least above some threshold.
Deviations from the predicted pattern point to problems with the integration
algorithm.
01 04 05 Default absolute tolerance for LL maximisation increased from
0.00001 to 0.001
01 04 05 Corrected label on LL tolerance in ML outputheader from 'Fractional'
to 'Absolute'
Major bugs fixed
30 03 05 Invalid FP operation when 'Auto initial values' selected
with 'Uniform' detection model
31 03 05 Zippin closed-population estimator failure criterion (Seber 1982 p.502)
09 04 05 SpotSim bug fixed (valid requests previously gave access violations)
Minor bugs fixed
18 03 05 Fixed bug in Poweranalysis : suppress output of open-population
header when turnover disabled
18 03 05 Fixed bug in Poweranalysis : Now uses correct title line for simulation
output (i.e. from PowerAnalysis page in Options)
10 04 05 Ctrl-T (Transfer to PowerAnalysis) bugs fixed
10 04 05 Minor TrapBuilder bug (colour of traps as dots) fixed
10 04 05 Bug in loadfile fixed
10 04 05 Bug in selection of output fields fixed
Minor improvements to interface
22 03 05 Output header for Distance analyses now records the setting of Monotonicity
22 03 05 Double click Session Selector on main form to return to the first
session
22 03 05 Display read-only initial values in memo field on main form; double-click
to change
02 04 05 Added check for traps in masked non-habitat
06 04 05 Power Analysis Log and Output tool buttons now enabled from the beginning
06 04 05 Power Analysis 'Trap success' field TS% now enabled from the beginning
10 04 05 Closure statistic moved within output fields to precede all spatial
statistics
11 04 05 Choice of MARK or plain format for export of capture histories
11 04 05 'Fields in output' button added to main form