Changes in 4.3 & 4.4
Changes apply from 4.4.1 except as indicated
Main form
- MMDM and ARL boundary-strip calculations now give intended values (4.3.1)- 'Number of recaptures' (NRecapt) now always includes recaptures within an occasion (as with a proximity detector, or when duplicates are expressly allowed)
- Huggins (linear logit) closed population estimator now accounts for losses on capture (note also change in default model)
- New graphic option on map popup menu 'Show detector usage', best used in combination with 'One occasion at a time' to display incomplete layouts
- 'Reset' now prompts for reset of all options
- Changes to habitat mask settings in Options take effect without new 'Read data'
- 'Tools | Export | Coordinates of mesh' now includes option to limit points to those within a specified distance of the nearest trap (4.4.2)
- added constant to log likelihood for closed population estimators Mh 2-point mixture and Mh beta-binomial for consistency with other closed-population estimators (4.4.6)
ML SECR (density estimation by maximum likelihood)
- automatic initial values: recognise invalid sigma and default to RPSV if necessary (4.3.2)- constant (multinomial) term now calculated consistently (this has no effect on estimates, but likelihood values may differ slightly from previous versions)
- trap layout files are now correctly matched with capture data when running a single session from among several
- improved display of results using 'fixed' link
- units of density as reported in the log under 'Fitted parameter' are now consistent with elsewhere (i.e. per km^2 if selected in Options|Output rather than default per ha)
- not all detectors need be used on all occasions - usage is defined with an optional extension to the trap layout format, and activated in Options | Input
- significant bugs fixed in handling of time and trap covariates, and extended reporting of scaled covariate values in the log
- model notation for 't' models changed to identify the actual time covariate (e.g. g0(T1) instead of g0(t))
- a different mesh may be imported for each session (Options | Computation) (4.4.2)
PowerAnalysis (simulation)
- 'Simulation stack' and 'Trapping and estimation stack' may be edited in a text window (4.3.1)- Some 'Advanced' 'trapping and estimation' options moved to a new page, and boundary-strip methods added to the 'Advanced' page (4.3.1)
- Attempt to calculate inferred strip width only when N-hat > 0 (4.3.2)
- Important change in default setting: no movement between occasions
- new output field NotRlsd for the number of animals dying in traps (if trap mortality > 0)
- bimodal halfnormal range simulation repaired, but remains undocumented (see Advanced options for proportion in second mode and distance between modes)
- export of captures now records deaths on capture as negative occasion number
- fixed reporting of randomly selected trap files
- some new export options for trap map (see popup menu) (4.4.5)
- Vart traps now works with 'Each session' (4.4.6)
Options
- New viewer for integration meshes (4.4.2)- Enhanced preview of vector habitat masks (4.4.2)
- Export clipped mask from Options | Habitat mask preview popup menu (4.4.2)
Minor improvements to interface
- check on input ('Read data') that number of sessions in capture data does not exceed maximum (200) (4.3.1)- ETA density calculator does not attempt variance calculation when perimeter is concave
- Output field 'NRecap' renamed to 'NRecapt'
- 'Losses on capture' now noted with session statistics on main form
- New Help button in Options, and help is more context-sensitive
- 'Tools | ML SECR log likelihood' minor improvements and correct area units (4.4.2)
- added check for ill-conditioned data for Mh jackknife estimate (4.4.6)
Change in default settings and limits
- 'Movement during sampling' defaults to None in PowerAnalysis (previously independent halfnormal)- Default model for Huggins (linear logit) closed population estimator is now the null model, not a learned response as previously (Options | Closed N)
- Increased precision convex trapping areas (default raised to 200 points/trap in Options | Miscellaneous | ETA perimeter points) (4.3.2)
- No specific limit to number of vertices in mask file (previously 20000) (4.4.2)
- Maximum number of occasions increased to 150 (4.4.4)
Closed population estimates
- check for illconditioned data Burnham & Overton jackknife (4.4.5)- constant added to log likelihood for 2-part mixture and beta-binomial estimates (no effect on estimates, but likelihoods and AIC now comparable for ML estimators) (4.4.5)
Other bugs fixed
- MMDM and ARL values corrected (4.3.1)- MLInfrW now works when different trap layouts are used with a between-session model (4.3.2)
- Options | Miscellaneous | ETA perimeter points not ignored in calculation of ETA (4.3.2)
- Program CAPTURE interface does not crash on first usage
- session-specific habitat masks now working (4.4.2)
- fixed bug that could cause Density to lock up when there were both few recaptures and some losses on capture (4.4.2)
- corrected apparent bug in likelihood for Pradel's gamma (4.4.6)