WiBB: an integrated method for quantifying the relative importance of predictive variables

Authors: Li, Qin; Kou, Xiaojun

(Ecography；DOI: 10.1111/ecog.05651)

• full_datasets: including independent datasets for three correlation structures, each with 200 datasets. Each dataset has a sample size of 1000, with one response variable (y) and four predictors (x1, x2, x3, x4). The first three (genuine) predictors were set to be strongly, moderately, and weakly correlated with the response variable, respectively (denoted by large, medium, and small Pearson correlation coefficients, r), while the correlation between the response and the last (spurious) predictor was set to be zero.

  • dt_lm_r1 (correlation structure: r = 0.3, 0.2, 0.1, 0.0);
  • dt_lm_r2 (correlation structure: r = 0.6, 0.4, 0.2, 0.0);
  • dt_lm_r3 (correlation structure: r = 0.8, 0.6, 0.3, 0.0);

• sub_datasets: resampled datasets with varying sample sizes from the full datasets for fitting LM and GLM separately;

  • dataset_lm: R data (.rds) for three correlation structures and six sample sizes (25, 50, 100, 200, 500, 1000);
  • dataset_glm: R data (.rds) for three correlation structures and six sample sizes (25, 50, 100, 200, 500, 1000);

• fit_result: example LM and GLM fitting results with the sample size = 1000;

  • r1_s6_lm_res.rds: an example fitting result by linear model, including estimates of variable importance based on sum of weight (sw), relative sum of weight (s_wi), standardized beta regression (b_star), WiBB (wi), and model ranking of the full candidate model set (ms.rank);
  • r1_s6_glm_res.rds: an example fitting result by generalized linear model; same content as the above;

• empirical_dataset: Mimulus data

  • mimulus_occ_var.csv: 71 Mimulus species occurrence coordinates and six associated climatic variables (temperature of cold- est month, T_cold; growing degree days above 0 Celsius degree, GDD0; precipitation seasonality, P_season; aridity of growing season, Aridity; isothermality, ISO; and temperature-precipitation synchronicity, TP_syn);
  • background_pts_var.csv: background locality data (randomly sampled from the overall distribution) and associated climatic data (same climatic variables as the above);

• Rscripts_0_functions.R: custrom functions, including data simulation, data sampling, and counting the correct ranking frequency;
• Rscripts_1_data_simulation.R: data simulation and sub-sampling for later LM and GLM;
• Rscripts_2_analysis.R: LM and GLM model fitting and calculations of variable importance by four indices: SW, SWi, beta*, WiBB, with varying sample sizes and bootstrap replicates;
• Rscripts_3_empirical_dataset.R: apply WiBB method to an empirical dataset of 71 Mimulus species;