edits for ms

Author: EricArcher

DESCRIPTION

@@ -7,7 +7,7 @@ Description: Estimate significance of importance metrics
     metrics for each predictor variable and p-value of
     observed. Provides summary and visualization functions for 'randomForest' 
     results.
-Version: 2.1.2
+Version: 2.1.5
 Authors@R: c(
   EA = person("Eric", "Archer", email = "[email protected]", role = c("aut", "cre")))
 BugReports: https://github.com/EricArcher/rfPermute/issues

NAMESPACE

@@ -4,14 +4,14 @@ S3method(plot,rp.importance)
 S3method(rfPermute,default)
 S3method(rfPermute,formula)
 export(classConfInt)
-export(clean.rf.data)
+export(cleanRFdata)
 export(confusionMatrix)
-export(exptd.err.rate)
+export(exptdErrRate)
 export(impHeatmap)
 export(pctCorrect)
 export(plotNull)
 export(plotVotes)
-export(proximity.plot)
+export(proximityPlot)
 export(rfPermute)
 export(rfPermute.default)
 export(rfPermute.formula)
@@ -23,11 +23,15 @@ importFrom(ggplot2,aes_string)
 importFrom(ggplot2,coord_flip)
 importFrom(ggplot2,element_blank)
 importFrom(ggplot2,element_rect)
+importFrom(ggplot2,facet_wrap)
 importFrom(ggplot2,geom_bar)
+importFrom(ggplot2,geom_density)
+importFrom(ggplot2,geom_histogram)
 importFrom(ggplot2,geom_point)
 importFrom(ggplot2,geom_polygon)
 importFrom(ggplot2,geom_raster)
 importFrom(ggplot2,geom_rect)
+importFrom(ggplot2,geom_vline)
 importFrom(ggplot2,ggplot)
 importFrom(ggplot2,ggtitle)
 importFrom(ggplot2,guide_colorbar)
@@ -39,9 +43,6 @@ importFrom(ggplot2,xlab)
 importFrom(ggplot2,ylab)
 importFrom(grDevices,chull)
 importFrom(grDevices,rainbow)
-importFrom(graphics,abline)
-importFrom(graphics,par)
-importFrom(graphics,plot)
 importFrom(gridExtra,grid.arrange)
 importFrom(parallel,detectCores)
 importFrom(parallel,makeForkCluster)
@@ -54,7 +55,6 @@ importFrom(reshape2,melt)
 importFrom(stats,binom.test)
 importFrom(stats,cmdscale)
 importFrom(stats,complete.cases)
-importFrom(stats,density)
 importFrom(stats,model.frame)
 importFrom(stats,model.response)
 importFrom(stats,na.fail)

R/clean.rf.data.R renamed to R/cleanRFdata.R

@@ -14,7 +14,7 @@
 #' @importFrom stats complete.cases
 #' @export
 #' 
-clean.rf.data <- function(x, y, data, max.levels = 30) {
+cleanRFdata <- function(x, y, data, max.levels = 30) {
   data <- as.data.frame(data)
   if (is.null(colnames(data))) colnames(data) <- 1:ncol(data)
   x <- setdiff(x, y)

R/confusionMatrix.R

@@ -9,7 +9,7 @@
 #' 
 #' @author Eric Archer \email{eric.archer@@noaa.gov} 
 #' 
-#' @seealso \code{\link{classConfInt}}
+#' @seealso \code{\link{classConfInt}}, \code{\link{exptdErrRate}}
 #' 
 #' @examples
 #' data(mtcars)
@@ -26,7 +26,7 @@ confusionMatrix <- function(rf, conf.level = 0.95, threshold = 0.8) {
   # Get confidence intervals
   ci <- classConfInt(rf, conf.level = conf.level, threshold = threshold)
   # Get expected error rate (prior)
-  prior <- exptd.err.rate(rf)
+  prior <- exptdErrRate(rf)
   prior <- (1 - prior[c(2:length(prior), 1)]) * 100
   # Add rows and columns
   conf <- rbind(conf, Overall = rep(NA, ncol(conf)))

R/dataSets.R

@@ -0,0 +1,12 @@
+#' @docType data
+#' @name symb.metab
+#' @title Symbiodinium type metabolite profiles
+#' @description A data.frame of 155 metabolite relative concentrations for 
+#'   64 samples of four Symbiodinium clade types.
+#' @usage data(symb.metab)
+#' @format data.frame
+#' @references Klueter, A.; Crandall, J.B.; Archer, F.I.; Teece, M.A.; Coffroth, M.A.	
+#'   Taxonomic and Environmental Variation of Metabolite Profiles in Marine 
+#'   Dinoflagellates of the Genus Symbiodinium. Metabolites 2015, 5, 74-99.
+#' @keywords datasets
+NULL

R/exptd.err.rate.R renamed to R/exptdErrRate.R

@@ -1,16 +1,16 @@
 #' @title Expected Error Rate
-#' @description Calculate expected OOB error rates for randomForest 
-#'   classification model based on random assignment and class sizes (prior).
+#' @description Calculate expected OOB error rates (priors) for randomForest 
+#'   classification model based on random assignment and class sizes.
 #' 
 #' @param rf an object inheriting from \code{link{randomForest}}.
 #' 
-#' @return a vector of expected error rates for each class.
+#' @return a vector of expected error rates (priors) for each class.
 #' 
 #' @author Eric Archer \email{eric.archer@@noaa.gov}
 #' 
 #' @export
 #' 
-exptd.err.rate <- function(rf) {
+exptdErrRate <- function(rf) {
   if(!inherits(rf, "randomForest")) {
     stop("'rf' is not a randomForest or rfPermute object.")
   }

R/plotNull.R

@@ -5,14 +5,19 @@
 #' call to \code{\link{rfPermute}}.
 #' 
 #' @param x An object produced by a call to \code{\link{rfPermute}}.
+#' @param preds a character vector of predictors to plot. If \code{NULL}, then 
+#'   all predictors are plotted.
 #' @param imp.type Either a numeric or character vector giving the 
 #'   importance metric(s) to plot.
 #' @param scale Plot importance measures scaled (divided by) standard errors?
-#' @param ... Optional graphical arguments to be sent to \code{\link[graphics]{par}}.
+#' @param plot.type type of plot to produce: \code{"density"} for smoothed density 
+#'   plot, or \code{"hist"} for histogram.
 #' 
-#' @details The function will generate an individual plot for
-#'   each variable and importance metric on the default graphics
-#'   device.
+#' @details The function will generate an plot for each predictor, with facetted 
+#'   importance metrics. The vertical red line shows the observed importance 
+#'   score and the p-value is given in the facet label. 
+#'   
+#' @return A named list of the \code{ggplot} figures produced is invisbly returned.
 #'   
 #' @author Eric Archer \email{eric.archer@@noaa.gov}
 #' 
@@ -22,44 +27,67 @@
 #'   ozone.rfP <- rfPermute(Ozone ~ ., data = airquality, ntree = 100, na.action = na.omit, nrep = 50)
 #'   
 #'   # Plot the null distributions and observed values.
-#'   layout(matrix(1:6, nrow = 2))
 #'   plotNull(ozone.rfP) 
-#'   layout(matrix(1))
 #' 
-#' @importFrom graphics abline par plot
-#' @importFrom stats density
+#' @importFrom reshape2 melt
+#' @importFrom ggplot2 ggplot aes_string geom_histogram geom_density xlab
+#'   ggtitle geom_vline facet_wrap
 #' @export
 #' 
-plotNull <- function(x, imp.type = 1, scale = TRUE, ...) {
+plotNull <- function(x, preds = NULL, imp.type = NULL, scale = TRUE, 
+                     plot.type = c("density", "hist")) {
+  
   if(!inherits(x, "rfPermute")) stop("'x' is not of class 'rfPermute'")
-  imp <- randomForest::importance(x, scale = scale)
-  imp <- imp[, c(ncol(imp) - 1, ncol(imp))]
+  imp <- randomForest::importance(x, type = NULL, class = NULL, scale = scale)
 
+  if(is.null(imp.type)) imp.type <- colnames(imp)
+  imp.type <- unique(imp.type)
   if(is.character(imp.type)) {
    not.found <- imp.type[!(imp.type %in% colnames(imp))]
    if(length(not.found) > 0) {
      imp <- paste(not.found, collapse = ", ")
      stop(paste("imp.type: ", imp, " is not in 'x'", sep = ""))
    }
   } else if(is.numeric(imp.type)) {
+    imp <- imp[, c(ncol(imp) -1, ncol(imp))]
     if(!all(imp.type <= ncol(imp))) stop("some 'imp.type' out of range")
     imp.type <- colnames(imp)[imp.type]
   } else stop("'imp.type' is not a character or numeric vector")
 
   sc <- if(scale) "scaled" else "unscaled"
 
-  op <- par(..., no.readonly = TRUE)
-  for(p in rownames(imp)) {
-    for(i in imp.type) {
-      n <- x$null.dist[[sc]][p, i, ]
-      o <- imp[p, i]
-      xlab <- if(is.character(i)) i else colnames(imp)[i]
-      pval <- x$pval[p, i, sc]
-      main <- c(paste("Variable:", p), 
-                paste("P(null >= obs) =", sprintf("%0.3f", pval)))
-      plot(density(n), xlim = range(c(n, o)), xlab = xlab, main = main)
-      abline(v = o, lwd = 2)
-    }
+  if(is.null(preds)) preds <- rownames(imp)
+  preds.not.found <- setdiff(preds, rownames(imp))
+  if(length(preds.not.found) > 0) {
+    not.found <- paste(preds.not.found, collapse = ", ")
+    stop(paste("The following predictors could not be found:", not.found))
   }
-  par(op)
+  
+  plot.type <- match.arg(plot.type)
+  g <- sapply(preds, function(p) {
+    df <- melt(
+      sapply(imp.type, function(i) x$null.dist[[sc]][p, i, ]),
+      value.name = "importance",
+      varnames = c("rep", "imp.type")
+    )
+    obs <- melt(
+      imp[p, imp.type, drop = FALSE], 
+      value.name = "importance",
+      varnames = c("predictor", "imp.type")
+    )
+    
+    pval <- x$pval[p, imp.type, sc]
+    labels <- paste0(names(pval), " (p = ", sprintf("%0.3f", pval), ")")
+    levels(df$imp.type) <- levels(obs$imp.type) <- labels
+    
+    pl <- ggplot(df, aes_string("importance"))
+    pl <- pl + if(plot.type == "hist") geom_histogram() else geom_density()
+    pl <- pl + xlab("Importance") + ggtitle(p)
+    pl <- pl + geom_vline(aes_string(xintercept = "importance"), color = "red", data = obs) 
+    pl <- pl + facet_wrap(~imp.type, scales = "free")
+      
+    print(pl)
+  }, simplify = FALSE, USE.NAMES = TRUE)
+  
+  invisible(g)
 }

R/proximity.plot.r renamed to R/proximityPlot.R

@@ -30,14 +30,14 @@
 #' data(mtcars)
 #' rf <- randomForest(factor(am) ~ ., mtcars, proximity = TRUE)
 #' rf
-#' proximity.plot(rf)
+#' proximityPlot(rf)
 #' 
 #' @importFrom stats cmdscale
 #' @importFrom grDevices chull rainbow
 #' @importFrom ggplot2 ggplot aes geom_point labs theme element_blank geom_polygon element_rect
 #' @export
 #' 
-proximity.plot <- function(rf, dim.x = 1, dim.y = 2, 
+proximityPlot <- function(rf, dim.x = 1, dim.y = 2, 
                            legend.loc = c("top", "bottom", "left", "right"),
                            point.size = 2, circle.size = 8, circle.border = 1, 
                            hull.alpha = 0.3, plot = TRUE) {

R/rfPermute.R

@@ -31,7 +31,7 @@
 #' \code{\link{plotNull}} for plotting null distributions from the \code{rfPermute} objects. \cr
 #' \code{\link{rp.importance}} for extracting importance measures. \cr
 #' \code{\link{rp.combine}} for combining multiple \code{rfPermute} objects.\cr
-#' \code{\link{proximity.plot}} for plotting case proximities.\cr
+#' \code{\link{proximityPlot}} for plotting case proximities.\cr
 #' \code{\link{impHeatmap}} for plotting a heatmap of importance scores.\cr
 #' \code{\link{randomForest}}
 #'

README.md

@@ -27,6 +27,7 @@ To install the latest version from GitHub:
 ```r
 # make sure you have Rtools installed
 if (!require('devtools')) install.packages('devtools')
+
 # install from GitHub
 devtools::install_github('EricArcher/rfPermute')
 ```
@@ -40,23 +41,26 @@ devtools::install_github('EricArcher/rfPermute')
 ## Current Functions
 
 `classConfInt` Classification Confidence Intervals  
-`clean.rf.data` Clean Random Forest Input Data  
+`cleanRFdata` Clean Random Forest Input Data  
 `confusionMatrix` Confusion Matrix  
-`exptd.err.rate` Expected Error Rate  
+`exptdErrRate` Expected Error Rate  
 `impHeatmap` Importance Heatmap  
 `pctCorrect` Percent Correctly Classified  
 `plotNull` Plot Random Forest Importance Null Distributions  
-`plot.rp.importance` Plot Random Forest Importance Distributions  
 `plotVotes` Vote Distribution  
-`proximity.plot` Plot Random Forest Proximity Scores  
+`plot.rp.importance` Plot Random Forest Importance Distributions  
+`proximityPlot` Plot Random Forest Proximity Scores  
 `rfPermute` Estimate Permutation p-values for Random Forest Importance Metrics  
 `rp.combine` Combine rfPermute Objects  
 `rp.importance` Extract rfPermute Importance Scores and p-values  
 
-## version 2.1.2 (current on GitHub)
+## current on GitHub (version 2.1.5)
 
 * Added `type` argument to `plotVotes` to choose between area and bar charts.
 * Changed plot.rfPermute to plotNull to avoid clashes and maintain functionality of plot.randomForest.
+* Changed name of `proximity.plot` to `proximityPlot`,  `exptd.err.rate` to `exptdErrRate`, and `clean.rf.data` to `cleanRFdata` to make camelCase naming scheme more consistent in package.
+* Changed `plotNull` from base graphics to ggplot2.
+* Added `symb.metab` data set.
 
 ## version 2.1.1