New functions and function improvements

[sjevelazco]

New Functions

• Explore variable selection based on univariate model performance, see Oeser et al., 2024
• New function to calculate inform user about the last update made in repository.
• New function to calculate variable importance. @mrose048
• New function to perform null models. @AdmirJr @sjevelazco
• New function or adapt sample_background or pseudo-absences function to perform stratified sampling @iminoli-dev
• New function to calculate range change/exposure based on continuous or binary maps and can produce an output with corresponding emissions/ssp and gcm and algorithm @mrose048
• New function to calculate ensemble of different GCM for different SSP @sjevelazco (use this function in the new vignette)
• New function to perform ensemble of ESM. @sjevelazco
• New function to select occurrences filtered with several values (occfilt_env and occfilt_geo) and return the best based on Velazco et al 2020 approach. @sjevelazco

Addapt implemented functions

• Addat function for partial dependence plot of ESM. Partial dependence plots for ensemble models #335 @sjevelazco
• Improve fit_raf and tune_raf to use or test (tune) different numbers of trees (based on adm package) @sjevelazco
• Improve fit_net and tune_net (ask Admir) @AdmirJr& @sjevelazco
• adapt correct_colinvar() to perform person and vif only using environmental condition of points (presences and absences) (think more about it if it worth implement this also for pca and fa approaches) @sjevelazco
• add a 'performance_part' (model performance for each partition and replica) as part of the outputs in each fit_ tune_ esm_ and ensemble_ functions. See implementation in adm. Possibility to extract performance metrics per cross-validation fold #312 @sjevelazco
• Adapt occfilt_env() to test several bins. @sjevelazco
• Adapt occfilt_geo() to test several values for all approaches @sjevelazco
• In sdm_eval() implement new metrics (see shared FUTURE functions and improvements.rtf in our shared Dropbox folder).
• adapt flexsdm to deal with temporal explicit data. Start with geographical and environmental filters @sjevelazco
• Write a code to calculate a default value for the parameter size in the net method. It must be implemented in fit_net. See the structure of 'mtry' argument in fit_raf function to guide you. (example here)
• PCA performed for calibration areas.

Vignettes

• Write a vignette to show an entire SDM workflow for several species @mrose048
• Vignette to create hierarchical SDM @iminoli-dev
• Vignette to create SDM with temporal resolution @iminoli-dev #395

[flee598]

Hi, first, thank you for the fantastic package, I am finding it very useful!
I came here to request/suggest the addition of a function to calculate variable importance, I see it is already on the list above, so I just wanted to say thank you put my vote towards such a function. Cheers, Finn.