knitr::opts_chunk$set(echo = TRUE)
The analysis consists of various small (drone) scripts. The order is indicated, but instead of calling them by hand, giving the model names as inputs, this script does this job for you, as the hive does the coordination job for the drones. (source for hive-drone terminology).
The conductor script here calls the drone scripts of the analysis in order to run the full analysis from one place.
The script vignettes/analysis_nonpublic.R
loads the necessary requirements, specifying the specific requirements by variable sections_to_be_loaded
. Each section includes an option to load the requirements.
Requires access to the Plant Ecology Group (Fischer) drive "planteco" at the IPS Bern. See header of the analysis_nonpublic.R
file.
ksource("vignettes/clean_and_load_soil_covariates.Rmd") ksource("vignettes/calc_covariates.Rmd") ksouce("vignettes/calc_betadiversities.Rmd") ksource("vignettes/function_imputation.Rmd") # Note : plotNAset is created here. # potentially need to re-run calc_betadiversities.Rmd # again with the correct number of plots!! # ksouce("vignettes/calc_betadiversities.Rmd") #TODO add early scripts
Scripts depending on output of each other, need to run per model.
Requirements
if(!exists("model_names")){ sections_to_be_loaded <- c() source("vignettes/analysis_nonpublic.R") }
Chose model by hand
# select a given model model_names_selection <- model_names[which(model_names$modelname == "gdm_EFturnover_0.7_LUI"), ] # # run analysis and create results ksource("vignettes/prepare_and_run_GDM.Rmd") # default model : EF turnover 0.7 LUI ksource("vignettes/check_GDM_input.Rmd") # produce nice plots/ results about correlations among input variables ksource("vignettes/plot_modelwise_GDM.Rmd") # produce bar and line plots of the current model
Loop through all models. Dont forget to transfer the output from vignettes/out to pathtodata
.
source("vignettes/analysis_nonpublic.R") # run all single function and multifunctionality models for(mod in model_names[model_class %in% c("multifun", "singlefun") & lui == "LUI", modelname]){ model_names_selection <- model_names[which(model_names$modelname == mod), ] # run analysis and create results ksource("vignettes/prepare_and_run_GDM.Rmd") ksource("vignettes/check_GDM_input.Rmd") # produce nice plots/ results about correlations among input variables ksource("vignettes/plot_modelwise_GDM.Rmd") # produce bar and line plots of the current model }
Run on cluster
# run cluster_scripts by hand on cluster! print("run cluster_scripts_documentation.Rmd by hand on cluster!")
Note : the below scripts only need to be run once, after the above scripts have been run for all models.
ksource("vignettes/check_GDM_diagnostics.Rmd") # 1 script for all model diagnostics ksource("vignettes/plot_unique_GDM.Rmd") # plot single EF models ksource("vignettes/summarise_GDM_results.Rmd") # create summary tables of all model results (deviance explained and isplines)
In order to calculate a summary model over all thresholds over all x values (not just maximum x, but a full line), the sd needs to be calculated everywhere.
Run the thresholds part of cluster_scripts_documentation.Rmd
on the cluster.
If no cluster is available : create a script "gdm_uncertainty.R". Copy paste the part from cluster_scripts_documentation.Rmd
in this file. Run it through all nestedness and turnover tresholds with LUI. Note that this script calculates sd for 1 model. In order to run through all, the input has been changed manually (you find code to change automatically in cluster_scripts_documentation.Rmd
).
source("gdm_uncertainty.R")
The output are 18 .Rds files for thresholds 0.1 - 0.9 for turnover and nestedness. E.g. "analysis/output_datasets/uncertainty_calc/gdm_EFnestedness_0.8_LUI_uncertainty.Rds"
ksource("vignettes/GDM_multifun_thresholds.Rmd") # calcluate models and generate plots (line and barplots) # could be added : a script plotting barplots with se
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.