-
corSelect
- added 'simplif' argument
-
distMat (affecting getRegion)
- method "auto": "haversine" only if is.lonlat(perhaps=TRUE), otherwise "euclidean"
-
getRegion
- 'prj=TRUE' used only for plotting; output always in input CRS
- distance matrix diagonal converted to NA before computing mean
- distances no longer squared-rooted (results better after 0-1 normalization)
- new 'decay' argument (default TRUE) if type="inv_dist"
- 'clust_dist' divided by 2 for clust_type ="buffer"
-
selectAbsences
- reclassify to zero and message (not error) if negative 'bias' values
- help file: add example bias layer with 'spatstat' kernel density, for target group background approach
-
Fav
- internally create object 'odds <- pred / (1 - pred)', to avoid repeated computation
- help file more explicit about 'model' being optional, and about training prevalence not being necessarily the same as in the raw input data
-
gridRecords
- specify (in both function and help file) that argument 'absences' (like 'abs.coords') can only be used if 'species' not provided (bug report by Alba Estrada)
- getRegion
- sort distance sums for type="inv_dist"
- CITATION
- add citation of package with self-updating version and year
- cleanCoords
- consider NAs as presences in 'abs.col'
- fix error when year.min = NULL
-
cleanCoords
- 'rm.zero' now executed before 'rm.equal'
- reordered some arguments
-
distMat
- argument 'dist_method' renamed to 'method'
-
distPres, getRegion
- import modEvA::range01() instead of re-defining it
- distPres.Rd, distMat.Rd, getRegion.Rd
- updated to reflect CRS assumption
- getRegion
- add 'prj' argument
- getRegion
- revert conversion of dist_mat upper triangle to NA (bug for type="inv_dist")
- for type="inv_dist", replace sapply() with rowSums() for much improved efficiency
- add 'col_reg' argument and change default polygon colour to gold
- getRegion
- check if CRS indeed empty before assigning EPSG:4326
- DESCRIPTION
- add BugReports field
- distMat.Rd, distPres.Rd, fuzzySim-package.Rd, selectAbsences.Rd, simMat.Rd
- 'dontrun' for examples that require setting CRS, to avoid check error on Windows old release
- distMat
- Sys.setenv("PROJ_LIB") to avoid check error on Windows old release (didn't work)
- corSelect
- new default 'cor' (more universal) for 'select' argument
- multicol
- new arguments 'vif.thresh', 'verbosity', 'plot', '...'
- can perform backward stepwise removal of variables based on VIF
-
getRegion
- use input 'dist_mat' also if clusters
-
corSelect
- add condition verbosity > 0 before most messages
- cleanCoords
- removed recently added transparency ('alpha') to point colours, which was not applicable to non-spatial inputs and slowed down the function for spatial inputs
-
getRegion
- in all cases (e.g. also for type='clust_width'), CRS assumed EPSG:4326 if perhaps is.lonlat()
- with type="clust_mean_dist", compute distances only within clusters (each less computer-intensive) rather than in the entire dist_mat
-
timer
- fixed new bug that made input expression be evaluated twice
- 'start.time' named argument replaced with '...' to work correctly on both POSIXct timestamps and expressions
- fuzzyRangeChange.Rd
- simplified some argument descriptions
-
timer
- input can now be an expression
-
cleanCoords
- reduced blue point size, so that red (removed) points are visible underneath
- added transparency to both point colours, to make point overlap visible
-
getRegion
- improved message about duplicate coordinates
-
gridRecords
- add na.omit() to match() call, to avoid NA error when 'species' is specified
-
getRegion
- reordered arguments: 'clust_dist' now before 'clust_type'
- use crds() instead of values() to fix false detection of duplicate coords
-
fuzzyRangeChange
- if input RasterLayer, coerce to SpatRaster to avoid obscure error downstream
- getRegion.Rd
- improved cluster description
- getRegion
- "clust_mean_dist" now gets cluster distances from subsets of the full distance matrix, instead of re-computing distances within clusters
- distMat
- now used by distPres() and getRegion()
-
getRegion
- new arguments 'clust_type', 'dist_method', and '...' for plot()
- distance now computed with distMat(), with several method options and "auto" default
- diagonal and lower triangle now removed before computing the mean of the distance matrix
- rows with no coordinates removed (with message) to prevent error
- message if duplicate coordinates found
-
distPres
- distance now computed with distMat()
-
multTSA
- now implemented also for SpatRaster + presence coordinates
-
multConvert
- now implemented also for SpatRasters
- removed loop
-
fuzzyRangeChange
- 'plot.type' default changed to new option "waterfall" (same as "barplot"), which seems the most appropriate for this
-
gridRecords
- some code changes to improve efficiency
-
multConvert.Rd
- mentions and examples for scale() and log() continuous variables
-
multicol.Rd, multConvert.Rd
- mention implementation also for SpatRasters
-
gridRecords.Rd
- example for gridding in tiles when 'rst' is large
- Fav
- added 'inv' argument to compute instead probability from favourability
- NEWS
- reformatted and converted to .md to work with news()
- biasLayer
- cleanCoords
- fixed bug when input data is SpatVector
- 'coord.cols' ignored with warning if provided and input is SpatVector
-
partialResp
- fixed bug when only one variable in model
- polygon instead of lines for confidence interval
- new argument se.mult, default 1.96 for the 95% confidence interval, but can be changed e.g. to 1 (for just the SE) or 0 (for no CI)
- initial plot with type="n" and points added if plot.points=TRUE, removing boilerplate code
- added args 'reset.par' and '...' for plot()
-
selectAbsences
- moved "Absences not enough..." section upwards, avoiding unnecessary remaining computations
-
distPres
- removed double loop to increase efficiency
- new default method "auto"
- option to use faster methods like haversine and cosine distance if packageVersion("terra") >= "1.8.7"
- partialResp, which can be used also for favourability (suggested by Alba Estrada)
- getRegion
- added 'dist_mat' argument
- changed default 'weight' to FALSE, as it looked too restrictive
- cleanCoords
- added 'extend' argument, default 0.1 to increase the output plot extent by 10% of the input coordinates range
- fuzzyOverlay
- fixed bug for raster inputs with NA pixels (reported by Jakub Nowosad: #2 (comment))
-
getRegion.Rd
- added examples
-
selectAbsences.Rd
- added example with bias layer
-
gridRecords
- output is SpatVector only if input is too
-
selectAbsences
- added data.frame() to fix error when input SpatVector
-
selectAbsences
- 'bias' can now be a raster layer
-
modelTrim
- implement categorical variables in phylolm models
-
fuzSim, simMat, fuzzyOverlay, modOverlap, fuzzyRangeChange
- inputs can be SpatRaster
-
simMat
- added 'plot' argument
-
fuzzyOverlay.Rd
- 'op' argument itemized list of options
-
NAMESPACE
- removed import of utils::installed.packages() [now using .packages(all.available = TRUE) instead]
-
getRegion
- labels now appear for all clusters in plot
-
gridRecords, selectAbsences
- output plot SpatVector when 'terra' available
- getRegion
- hclust() now with method = "single"
- cutree() with h = clust_dist (new argument)
- fixed bug in 'weight', which wasn't correctly ordered (replaced merge with match)
- 'weight' argument now also applies to type="width"
-
distPres, selectAbsences
- add 'CRS' argument; if provided, terra::distance() is used
-
distPres
- add 'verbosity' argument
-
getRegion
- fix plotting error, now specifying 'terra::' before plot() and text()
- getRegion
- modelTrim
- implemented for 'phylolm' models
- added 'phy' and 'verbosity' arguments
- changed 'if (class(model) %in% c("glm", "lm"))' to 'if (all(class(model) %in% c("glm", "lm")))', otherwise CRAN check error
- DESCRIPTION
- added contributors
- dms2dec
-
FDR, multGLM, corSelect
- added argument 'test' (default "Chisq" for back-compatibility; future default probably "Rao" score as in SPSS)
-
stepByStep
- message and NA outputs if model has no variables
- stepByStep
- added option select="p.value" and associated arguments
- stepwise.Rd
- changed 'sp.col' in Examples, for a model where a variable is included and then dropped
-
appendData
- 'data1' and 'data2' coerced to data.frame (to work e.g. for SpatVectors)
-
stepByStep
- 'data' can now be a 'glm' model object, from which the order of the variables will be taken
- selectAbsences
- added 'dist.mat' argument for 'distPres'
- updated help file to reflect this
- distPres.Rd
- corrected typo "euclidian" -> "euclidean"
- mention that (external) spherical distance is recommended for large extents
- mention that 'method' is only used if 'dist.mat' is not provided
- corSelect
- fixed incorrect use of all.equal() to compare results between different 'select' criteria (thanks to bug report by Jose Carlos Guerrero)
-
corSelect.Rd
- mention {collinear} pkg under "See also"
- mention that method 'pearson' is recommended for >30 rows, 'spearman' for >10 rows (as in ?collinear::collinear)
-
fuzSim.Rd
- mention fuzzy Jaccard = weighted Jaccard (Ioffe 2010), overlap, coincidence, consistence (Real et al. 2010)
-
corSelect
- added 'cor' to 'select' criteria
-
getPreds
- replaced 'requireNamespace(raster)' with 'raster %in% rownames(installed.packages())', to avoid loading the pkg if unnecessary
==
==
- cleanCoords
- added 'year.min' and 'year.na.pass' arguments
==
==
-
appendData
- added 'add.source' argument
- x and y arguments renamed to data1 and data2
- warning when no matching columns
- out data frame created empty, instead of taking data1 values and then converting to NA
-
gridRecords
- replaced 'requireNamespace(raster)' with 'raster %in% rownames(installed.packages())', to avoid loading 'sp' and triggering all the deprecation warnings
- cleanCoords.Rd
- now states that row names are carried from input dataframe
==
==
- appendData
==
==
- cleanCoords
- added 'terra::' before plot() on SpatVector objects, to prevent newly appeared error
-
Fav.Rd
- now states that 'model' or 'pred' must have been obtained with weights=NULL
-
corSelect.Rd
- added "however" references against dropping correlated variables
==
==
- multGLM
- added 'coeff' argument to pass to 'corSelect', and changed 'cor.thresh' default accordingly
- precluded error when no variables pass corSelect
==
==
- corSelect
- fixed bug that didn't allow correctly reporting the selected variables when coeff=FALSE
==
==
- corSelect
- added 'coeff' argument; if FALSE, high correlations are identified based on p-value (not coefficient) of correlation
- 'p.value' column added to 'high.correlations' output table
- fixed bug when there's >1 categorical variable to exclude a priori
==
==
- multGLM
- warning emitted if select="p.value" and trim=TRUE
- started implementation of 'block.cols' argument (still not useable)
==
==
- corSelect
- 'select' default now VIF if 'sp.cols' is null
- 'select' can now be null
- message displayed saying what 'select' criterion was used
- help file updated accordingly
==
==
-
cleanCoords
- added arguments 'rm.abs' and 'abs.col'
- when all records are removed, return 'data' with zero rows instead of error
- input 'data' can also be a SpatVector of points, in which case output is too
- 'coord.cols' default now NULL, can be taken from 'data' if class 'SpatVector'
- 'coord.cols' can be NULL, in which case only the other removal criteria (absences, coordinate uncertainty) are applied
-
gridRecords
- input 'pres.coords' and 'abs.coords' can also be SpatVector of points
==
==
- CITATION
- removed old-style personList(), and replaced citEntry() with bibentry(), as per new CRAN requirements
-
selectAbsences
- added 'df' argument
- output df now ordered by as.integer(rownames) rather than character rownames (1, 10, 100, 2, ...)
- points now plotted in this order: unselected absences, selected absences, presences on top
-
gridRecords
- points now plotted in this order: absences, presences on top
- cleanCoords
- added 'plot' argument
- changed default 'uncert.limit' from Inf to 50000
- selectAbsences
- cleanCoords
- fixed bug in 'rm.imprecise'
- cleanCoords
-
fuzzyRangeChange
- added argument 'plot.type'
- default 'plot.type' now "lollipop"
-
entropy
- added argument 'plot.type'
- default 'plot.type' now "lollipop"
- pairwiseRangemaps
- replaced 'PBSmapping' (and consequently 'sp' and 'maptools') with 'terra' (future-proof and much faster)
- DESCRIPTION
- packages 'PBSmapping', 'sp' and 'maptools' removed from 'Suggests'
-
gridRecords
- added 'plot' argument
-
corSelect
- non-numeric variables excluded with warning (instead of causing error)
-
multGLM
- corrected bug that miscounted the variables excluded by FDR when FDR.first=FALSE (affecting only console messages)
- total N variables now reported when TSA=TRUE (i.e. "... 46 with the spatial trend variable" instead of just "...plus the spatial trend variable")
-
modelTrim
- added family <- family(model) when method = "summary", as per bug report by J.C.Guerrero email 30/12/2022
- multGLM
- added 'FDR.first' argument
-
multGLM.Rd
- added recommendation to set trim=FALSE when select="p.value"
- added JCG funding acknowledgment to Note
-
fuzSim.Rd
- help file now mentions Jaccard and Sorensen also as recommended metrics for model evaluation
-
corSelect
- added 'verbosity' argument
- message emitted saying how many vars selected and excluded (like in 'FDR') if verbosity > 0
- message emitted saying which vars selected and excluded if verbosity > 1
-
FDR
- argument 'verbose' replaced with 'verbosity' (with warnings)
- message emitted saying which vars selected and excluded if verbosity > 1
-
multGLM
- argument 'verbose' for 'FDR' replaced with 'verbosity' (as per 'FDR' change above)
- argument 'select' now accepts additional criterion "p.value", in which case 'stepwise' used instead of 'step'
- added arguments test.in, test.out, p.in and p.out to pass to 'stepwise'
- trim=TRUE reverts back to using 'modelTrim' (rather than 'stepwise' with direcction='backward' as in previous experimental versions)
- argument trim.fun (added in previous experimental versions) removed
-
summaryWald
- fixed bug when model has no variables
-
stepwise
- break loop when a variable simultanously meets the criteria for inclusion and exclusion
- stepwise
- help file with new references (two against AIC, and one more against stepwise)
- gridRecords
- moved if(na.rm) uppward to go back to using 5:ncol(result) instead of names(result), which was causing problems with some raster names (bug report by Alba Estrada)
- ordered result by 'cell' and removed (confusing) row names
- help file Examples now includes mapping multi-species gridded presences
- multGLM
- help file notes that 'TSA' uses "type="Y" and is included in FDR, corSelect, etc.
- gridRecords
- warning if 'species' (not character or) has leading/trailing spaces
- added check.names=FALSE to data.frame call, so that output column names are exactly as the unique values of 'species' (even if they are numbers), and the first species name doesn't come out different from the rest
- added last argument in abs.coords <- terra::crds(rst, df=TRUE, na.rm=FALSE), otherwise some rows with missing data in some variables would be missing from the result (bug report by Alba Estrada)
- multGLM
- help file notes that 'modelTrim' and 'stepwise' have different default significance thresholds, to explain why 'stepwise' may leave more variables in the model
-
multGLM
- fixed message about N variables excluded by 'modelTrim' (now by trim.fun)
- set trace=0 for 'stepwise'
-
fuzzyConsensus
- changed default biplot to FALSE, as it makes function much slower
-
fuzzyRangeChange
- changed "Stable presence" and "Stable absence" to "Stable positive" and "Stable negative" to avoid erroneous interpretations
- added y axis label
-
stepwise
-
summaryWald
-
fuzzyRangeChange
- added 'x.lab=TRUE' argument, so user can set to FALSE and add labels differently
-
stepByStep
- added warning("'Favourability' is only applicable when family=binomial(link='logit'), so it was automatically set to FALSE.")
-
multGLM
- added argument trim.fun="modelTrim", which can be changed to "stepwise"
- stepByStep
- added 'k' argument to pass to 'step' (allows approaching p<0.05; https://www.researchgate.net/post/Why-stepAIC-gives-a-model-with-insignificant-variables-in-the-summarymodel)
- added 'direction' argument to pass to 'step' (implementation by Alba Estrada)
-
fuzzyConsensus
- 'cat' message now mentions the number of input vectors
-
multGLM
- added 'cor.method' argument (for 'cor' via 'corSelect'; default "pearson")
-
stepByStep
- 'sp.col' and 'var.cols' can now be column names, not just index numbers
- fuzzyConsensus
- 'data' can also be SpatRaster
- removed unnecessary argument 'fav.cols' (as no other '*.cols' argument exists)
- fuzzyConsensus
- Fav
- stops with error if any 'pred' is outside the [0,1] interval (matching 'fuzzyConsensus')
- DESCRIPTION
- added reference about the method, as suggested by Uwe Ligges after previous CRAN submission
- modOverlap
- improved help file and removed reference to requiring GLM predictions
- prevalence (copied from 'modEvA' pkg)
-
bioThreat
- "white" changed to "grey" for 0-values to be visible on the map
-
corSelect, FDR, fuzzyOverlay, integerCols, multConvert, multGLM, multicol, multTSA, simMat, splist2presabs, stepByStep, transpose
- added data=as.data.frame(data), to avoid errors when input is tibble
-
bioThreat, Fav, favClass, fuzSim, fuzzyRangeChange, modOverlap, sharedFav, spCodes
- added 'unlist' to avoid obscure error when input is one-column tibble instead of vector
- sharedFav
- improved help file and corrected a mistake (bug report by Alba Estrada)
- replaced 'main' argument with '...' for barplot (to accept also 'las')
- plot now produced in phases as computations are made, to have something plotted even in case of error
- 'conf' can be set to NA, to avoid error and still get values and plot (without CIs)
- returned value now includes 'FOvI' and 'bins_table'
-
rarity
-
vulnerability
-
entropy
-
spCodes
- added 'verbosity' argument
- 'cat' now used instead of red 'message' when OK no dupl codes
-
corSelect
- replaced '...' with arguments 'use' and 'method', the former defaulting to "pairwise.complete.obs"
-
simMat
- added 'na.omit' to 'data' in 'stopifnot', to avoid error when NAs
-
sharedFav
- added 'bin_interval' argument, currently accepting '0.1' (the default, for back-compatibility) or 'quantiles'
- added Note to help file about possible error when overly small bins
-
gridRecords
- result now ordered by "cells" column, instead of presences then absences
- help file "Examples" changed to "elev" raster for more clarity
- included also example with multiple species
-
corSelect
- added 'family' argument (to pass to FDR when 'sp.col' is not NULL)
- fixed NA 'excluded.vars' output when input 'var.cols' is character
-
FDR
- added 'Gamma' for positive non-integer responses when family="auto"
- gridRecords
- added 'species' argument for multi-species gridding
- 'pres.coords' (and 'abs.coords' if not NULL) automatically converted with 'as.data.frame'
- changed 'raster' to 'terra' in help file "Examples"
- fixed bug that did not grid abs.coords (if not NULL)
-
corSelect
- fixed new CRAN check error (with 'all.equal' in 'corSelect') on r-devel-linux-x86_64-fedora-clang and r-devel-linux-x86_64-fedora-gcc (by manually setting the 'tolerance' argument for 'all.equal')
-
Fav
- added import modEvA::mod2obspred
- fuzsim.Rd
- updated Sorensen wikipedia link with 'https'
-
Fav
- 'model' argument can now be of class 'glm', 'gam', 'gbm', 'randomForest' or 'bart'
- 'pred' can now also be a 'SpatRaster' ('terra' pkg) object
-
gridRecords
- 'rst' can now also be a 'SpatRaster' ('terra' pkg) object
- Raster* 'rst' coerced to SpatRaster if 'terra' pkg is installed
- FDR
- fixed bug when using 'pvalues' as input (following bug report by Stephen via modTools contact form)
- gridRecords
- fixed bug when only one raster layer [ , drop = FALSE]
-
simMat
- added 'verbosity' argument
- replaced "25%-50%-75% done" with 'utils::txtProgressBar'
-
multTSA
- added 'verbosity' argument
-
getPreds
- added 'verbosity' argument
- 'data' can now also be a RasterBrick (not only dataframe or a RasterStack)
-
Fav
- added 'verbosity' argument
- 'pred' can now also be a RasterLayer (not only a numeric vector)
- favClass (benefitting also bioThreat)
- added 'is.na(fav) | ' to 'stopifnot' to avoid NA fail
- added Linero et al. (2020) to references of papers using fuzzySim
- gridRecords
- added 'absences' logical argument
- gridRecords
- added 'na.rm' argument (default TRUE)
- fixed bug when no gridded absences (i.e. no absence AND not presence)
- clarifications and typo/format corrections in the manual
- gridRecords.Rd
- added examples
- gridRecords
- getPreds:
- replaced 'if("raster" %in% class(data))' (no longer working) with 'if (is(data, "RasterStack"))'
- added 'Imports: methods' to DESCRIPTION and 'importFrom("methods", "is")' to NAMESPACE
- if 'data' are raster, added 'raster::' before 'stack'
- multGLM:
- fixed id.col name in 'predictions' when id.col input as character
- added 'inst' folder with article citation information
- removed 'onAttach' function with citation message on load
- added package URLs to DESCRIPTION file
- multGLM:
- output now includes list of selected variables per model
- sp.cols, var.cols and id.col can now optionally be input as column names rather than index numbers
- fixed length of some lines along the PDF manual
-
modOverlap:
- corrected typo in na.rm closing parentheses (reported by Heidi K. Mod)
-
Fav:
- slightly reduced probabilities of exactly 1, which would cause division by zero (resulting Fav is still 1)
- sharedFav
- favClass
- bioThreat
-
multGLM:
- spatial_trend variable in models (when TSA=TRUE and the spatial trend is selected) now named after the response variable (e.g. 'sptrend_giraffe')
-
getPreds:
- fixed new bug by replacing 'if (class(data) == "RasterStack")' with 'if ("RasterStack" %in% class(data))'
- multGLM:
- added example with TSA=TRUE
- updated maintainer e-mail address
- multTSA:
- resulting model object now named 'models' rather than 'TSA.models'
- corrected bug in coordinate polynomial names (if save.models = TRUE)
-
distPres
- 'inv' now subtracts from 1 (after standardizing) rather than dividing by 1
-
multTSA:
- added 'criterion' argument, which can be 'AIC' (to use 'step') or 'significance' (to use 'modelTrim')
- added '...' argument to pass to modelTrim
- introduced more informative coordinate polynomial names (visible if 'save.models = TRUE')
- added 'simple = TRUE' to 'poly' for speedup
-
modelTrim:
- added 'data' argument to 'update', to avoid 'attach'
-
multGLM:
- added arguments 'verbosity', 'TSA' and 'coord.cols'
- added 'data = train.data' to 'step', and 'with(train.data' to 'model.formula', so finally got rid of 'attach'
- "Building model 1..." 'message' instances replaced with 'cat' so that they are saved if 'sink' is used
- multTSA:
- added argument 'type' which can be "Y", "P" (the default, for back-compatibility) or "F" (which substitutes the deprecated argument below)
- deprecated argument 'Favourability = FALSE'
-
multTSA:
- documented argument changes described above
-
pairwiseRangemaps:
- updated literature reference
- added additional article citing fuzzySim
- getPreds:
- 'data' can now be a RasterStack
- multTSA:
- fixed bug when only one species was used
- added articles citing fuzzySim
- corSelect:
- included 'VIF' criterion
-
multGLM:
- fixed bug when only one variabe passed to corSelect
-
corSelect:
- added check that 'data' is not missing and is a data frame
- help file now states that corSelect is included as option in multGLM
- added 'use = "pairwise.complete.obs"' to examples
-
PDF manual:
- reduced several code line lengths to avoid overboard
- fuzzyRangeChange:
- underscore replaced with space in measure names
- x axis labels plotted instead of legend
- 'col' no longer supplied by default
- pairwiseRangemaps:
- argument 'chunks' replaced with 'nchunks' for operative reasons
- added 'subchunks' argument for continuing interrupted runs
-
corSelect:
- removed error when no corrs above threshold
- sp.cols now NULL by default (to avoid error when missing)
-
multGLM:
- FDR correction reverted to "fdr" by default
-
FDR:
- BIC now also provided
-
corSelect:
- BIC now included as selection criterion
-
multGLM:
- added "select" argument (for 'step') - AIC or BIC
- multGLM:
- FDR correction now "BY" by default
-
corSelect:
- error message when length(sp.cols) > 1
- message about excluded missing data
- added option sp.cols=NULL to get only high.cor.mat
-
multGLM:
- added 'correction' argument to pass to 'FDR'
- fuzzyRangeChange:
- user can now choose which measures to calculate
- 'loss' now reported as originally negative
- results provided also as a barplot
- multTSA:
- eliminated call to 'attach'
-
pairwiseRangemaps (calculate area of pairwise intersection and union between rangemaps)
-
rangemapSim (calculate rangemap similarity using common similarity indices)
- FDR:
- now uses only finite sp.col values
-
fuzzyRangeChange:
- corrected proportional changes to be relative to reference range size, not total study area
- replaced "stable" with "stable presence" and added (fuzzy equivalent of) "stable absence"
- replaced "change" with "balance" (overall loss/gain, not amount of changed cells)
- result is now data frame, not named vector
-
distPres:
- now allows NA values
-
fuzzyOverlay:
- added arguments 'prop = TRUE' and 'overlay.cols = 1:ncol(data)'
- result now includes 4 (not 8) values, either sum or proportion
- result is now a named vector instead of a list
- 'na.rm' now TRUE by default
-
fuzSim:
- added argument 'na.rm = TRUE'
-
multGLM:
- bug corrected in corSelect (was reporting but not really eliminating variables)
- suppressMessages in corSelect
-
fuzzyOverlay: modifications reflecting function changes
-
modOverlap: example now provided
-
fuzzyOverlay, fuzzyRangeChange, modOverlap:
- examples containing 'jitter' corrected to avoid values outside the [0, 1] interval
-
fuzSim:
- examples now provided also for similarity between fuzzy data
- tables of significance for Jaccard and Baroni's indices now referred
-
fuzzyOverlay (calculate row-wise intersection, union, expansion, contraction or consensus among continuous model predictions)
-
fuzzyRangeChange (calculate overal loss, gain, and maintenance of favourability between models)
-
corSelect (select among correlated variables based on their bivariate relationship with the response)
-
modOverlap (asses the total overlap between model predictions using niche comparison metrics)
-
multGLM:
- 'corSelect' now included as additional option for variable selection
-
FDR:
- AIC now also calculated
- 'model.type' deprecated
- 'family = "auto"' by default
- 'simplif' argument added
-
multicol:
- 'model' argument added (user can provide a model object instead of a set of variables)
- variables in output are now ordered according to VIF
-
simFromSetOps:
- the similarity index used is now mentioned in a message
-
FDR:
- data input format changed, with former parameters 'response' and 'predictors' replaced with 'data', 'sp.cols' and 'var.cols' (for coherence and compatibility with 'multGLM' function)