Changes in version 1.2.1 - CITATION - added citation of paper and package with self-updating version and year - DESCRIPTION - added Date field for self-updating package citation Changes in version 1.2 (2025-12-09) - DESCRIPTION, NAMESPACE - removed dependency on the {collinear} package, as its updates were often not backward-compatible and introduced breaking changes - getModels() - collinearity is now reduced with fuzzySim::corSelect() and fuzzySim::multicol(), instead of collinear::collinear() Changes in version 1.1 (2025-07-07) - getPerformance() - added 'kappa' to implemented metrics, by reviewer request - create columns before they are populated, for efficiency - renamed and reordered some threshold columns, for better readability Changes in version 1.0 (2025-03-05) - replaced 'year' with 'period' in messages and outputs - following CRAN requests: - replaced cat() with message() conditional on 'verbosity' argument - replaced getwd() with tempfile() in Examples - replaced dontrun{} with donttest{} in examples that take long to run but are feasible without extra software; then replaced instead with if(interactive()), because check still runs the donttest{} section - removed LICENSE file for not having additional restrictions Changes in version 0.23 - getImportance() - added 'palette' argument - getModels() - periods (years) not automatically sorted, but rather taken as in input 'rasts' - help file: improved description of 'rasts' layer names Changes in version 0.22 - getImportance() - added 'plot' and associated arguments Changes in version 0.21 - getImportance() - new function Changes in version 0.20 - getTrend(), DESCRIPTION - replaced Kendall::MannKendall() with trend::mk.test() to reduce package dependencies; and eventually removed also mk.test, as slope and significance are sufficient and both are provided by trend::sens.slope() - full ouput includes just slope, upper and lower confidence limits, and p-value Changes in version 0.19 - getTrend() - added Sen's slope and associated statistics from trend::sens.slope() Changes in version 0.18 - getModels() - new default 10 (not 0) for 'nreps' - added collinear() new argument 'quiet=TRUE' after 'collinear' update - output has the same structure (list of lists) with or without replicates, for code consistency - getPredictions() - output is now SpatRasterDataset (not single SpatRaster if no replicates, or list of SpatRasters if replicates), for code consistency - fixed error when producing output file if replicates - argument 'mods' can be directly the output of getModels (not mods$models) - getPerformance() - output now includes both train and test performance, number of train and test presences, and TSS train and test thresholds - fixed code for plot=TRUE when replicates - argument 'data' replaced with 'mods' for safer use - aded 'sep.plots' and 'reset.par' arguments to modEvA::optiThresh() - getTrend() - implemented for replicates, converting SpatRasterDataset to SpatRaster and using mean prediction - DESCRIPTION - included additional authors - increased version dependence for collinear, to use non-default value for new argument 'quiet' - increased version dependence for modEvA, to use 'sep.plots' and 'reset.par' (new) argument values Changes in version 0.17 - getPerformance() - implemented output for replicates (still test performance only) Changes in version 0.16 - getPerformance() - added pbg = TRUE to modEvA functions (requires modEvA >= 3.20) - 'plot' default changed to FALSE, as it can be slow for large datasets - started implementing for models with replicates - name of first column of output changed to 'model' rather than 'year', to accomodate replicates Changes in version 0.15 - getPredictions() - implemented for models with replicates Changes in version 0.14 - getModels() - added 'bias' argument (requires fuzzySim >= 4.25) Changes in version 0.13 - getModels() - integrated test + nreps arguments in output 'data' - named list of replicate models Changes in version 0.12 - getModels() - implemented 'test' and 'nreps' arguments in output 'models' - help file mentions that constant variables and different collinearity structures may produce models with different sets of variables among periods - getPredictions() - currently stops if input 'mods' has replicates Changes in version 0.11 - getModels() - messages saying which variables removed by collinear() - constants checked before (rather than after) collin - added drop=FALSE to fix error when only one selected predictor Changes in version 0.10 - getModels() - improved code to check 'region' - README - some more additions for improved clarity Changes in version 0.9 - getTrend() - warning if no values produced due to small sample size - 'occs' can be of class 'SpatVector' or 'sf', in which case they aren't converted to data.frame and their CRS is taken - 'occs' projected (with message) if not same CRS as 'rasts' - README - some additions for improved clarity Changes in version 0.8 - getModels() - drop non-varying predictors (with message) to avoid maxnet() error - message if n pixels < nbg - getPredictions() - 'data' CRS set to the same as 'rasts', to avoid warning "[mask] CRS do not match" - added bibliographic reference for type="cloglog" recommended default - getRegion() - removed in favour of a more complete version which was already being programmed in package {fuzzySim} Changes in version 0.7 - getModels() - warning if input 'rasts' is not SpatRaster, which will work here but fail elsewhere Changes in version 0.6 - getPerformance() - new function - getModels() - output now includes three elements: $models, $replicates, $data (needed for getPerformance) - getTrend() - added arguments 'file', 'verbosity' - getVariables(), getModels(), getPredictions(), getTrend(), README - implemented 'file' including a folder path, e.g. "outputs/models" - folder created if not in getwd() - all functions - replaced instances of @import with @importFrom in roxygen skeleton Changes in version 0.5 - getModels() - added argument 'seed = 1' for the background sample - added arguments 'maxcor' and 'maxvif' to pass to collinear::collinear() - added argument 'classes' to pass to maxnet::maxnet.formula() - added argument 'regmult' to pass to maxnet::maxnet() - default 'nreps' currently 0 - added argument 'test = 0.2' for the test data proportion [STILL PENDING IMPLEMENTATION] - getPredictions() - added arguments 'clamp' and 'type' to pass to maxnet::predict() Changes in version 0.4 - getVariables() - changed url, following TerraClimate Batch Downloads instructions (https://www.climatologylab.org/wget-terraclimate.html) - set vsi=FALSE if not on unix platform, to prevent download fail - getVariables(), getModels(), getPredictions() - object imported from 'file', if it exists in getwd() Changes in version 0.3 - getTrend() - dealt with NA pixels to avoid numerous Kendall::MannKendall() warnings (thanks to Spacedman at https://gis.stackexchange.com/a/464198) - bonus: increased efficiency, as mannkendall outputs an unlisted vector (still thanks to Spacedman at https://gis.stackexchange.com/a/464198) Changes in version 0.2 - getPredictions() - added argument 'region' (to get predictions only within modelled polygon) - getTrend() - added argument 'occs' (to get trend only within presence pixels) - added Mann-Kendall outputs besides tau: p_value, S, S_variance Changes in version 0.1 - Added the following package functions, together with documentation and users guide: - getRegion() - varsAvailable() - getVariables() - pixelArea() - getModels() - getPredictions() - getTrend() Changes in version 0.0.0.9000 - Added a NEWS.md file to track changes to the package.