R/predicor.R

In wilsonfrantine/landscapeDecoupler: Landscape Decoupler: Tools for Multi-scale Landscape Analysis

#' @name predicor
#' @export
#' @title Predictor correllation throught scales
#' @description This function analyze the correlation of a predictor across different scales.
#' @return A data frame containing the correlation values between scales
#' @usage predicor(x, ...)
#' @param x a dataframe generated with the functions [calc_lsm] and tranformed with [lsm2multifit()] function
#' @param ... other parameters to be passed for the stats::cor function
#' @examples
#' \dontrun{
#' #Comparing correlation behaviour in nested and decouple multiscales:
#'
#' b <- c(500, 1000, 1500, 2000, 2500, 3000)
#'
#' nest.ls <- nestedscales(r, p, b)
#' dec.ls  <- decouple(r, p, b)
#'
#' #extracting percentage for each class and scale
#' nest.lsm <- calc_lsm(nest.ls, metric = "pland")
#' dec.lsm  <- calc_lsm(dec.ls,  metric = "pland")
#'
#' #Comparing correlation for the class = 3
#' nestcor <- predicor(lsm2multifit(nest.lsm, level="class", class=3,
#'  metrics="pland"))
#' deccor  <- predicor(lsm2multifit(dec.lsm, level="class", class=3,
#'  metrics="pland"))
#'
#' corcomp    <- data.frame(
#'   strategy    = factor(c(rep("nested", nrow(nestcor)),
#'                   rep("decoupled", nrow(deccor))),
#'                   levels = c("nested","decoupled")),
#'   scales      = factor(c(nestcor$scales, deccor$scales),
#'                   levels=nestcor$scales),
#'   correlation = c(nestcor$correlation, deccor$correlation)
#' )
#' library(ggplot2)
#' ggplot(corcomp, aes(x=scales, y=correlation, fill=strategy))+
#'   geom_bar(stat="identity", position="dodge")+
#'   theme(axis.text = element_text(angle=90))
#'
#'}

predicor <- function(x, ...){
 cor.v <- stats::cor(x[ , grepl(pattern ="X\\d+", colnames(x)) ], ...)
 scales <- t(utils::combn(colnames(cor.v), 2))
 comb <- paste(scales[,1], scales[,2], sep=".")
 data.frame(scales=factor(comb, levels=comb), correlation=cor.v[scales])
}