README.md

In snwikaij/GRASS: GRadient AnalySiS for ecologists

GRASS

Presence/absence data for taxa is often more prevalent than abundance. It is also more consistent and less variable than abundance data. For example, the abundance of fish schools or plant cover are very variable and notation of this factor is very dependent on sample effort and method. I started this R-package more for out of personal interest on how programming and data-analysis can quantify ecological phenomenon in stead than merely describe these. I also wanted to use methods that everyone could interpret easy even with minimal statistical background, the results: "GRASS".

The GRASS package has multiple functions. The basic of some functions relies on the Area Under the Curve (AUC). The AUC gives the probability random sample of taxon x will rank higher than a random sample from taxon y. The AUC is adapted so it can be used to analyse the variation of multiple taxa in an assembly in relation to a single gradient (i.e. after multivariate analysis). This can be used to answer the question, “how likely are we to find a specific taxon deviating from the other taxa (residual assembly).” The methods for this are presented as the AUC.test and AUCG.test functions.

The threshold indication functions are used to identify thresholds in presence/absence data. This function uses the cumulative sum principle to detect changes in presence/absence in relation to the sample distribution. These functions incorporate the AUC as selection criteria. The methods for this are presented as the cusTH, plotTH and plotCUS functions.

The random forest model function is a simplification of the functions used in the randomForest and party packages. It creates a training and validation dataset and produces accuracy measures with the caret and rel packages. Based on the total dataset it determines the relative importance of the used predictor variables.

snwikaij/GRASS documentation built on July 29, 2020, 1:54 p.m.