Merge pull request #10 from awsm-research/issue-9

Address issue #9

NAMESPACE

@@ -7,6 +7,7 @@ export(get.importance)
 export(loadDefectDataset)
 export(performance.calculation)
 export(plotVarClus)
+export(remove.constant.categorical)
 export(single.fit)
 export(stepwise.vif)
 import(dendextend)

R/AutoSpearman.R

@@ -12,7 +12,7 @@
 #' @param vif.threshold a numeric for a threshold of VIF score (default = 5)
 #' @param verbose  TRUE for printing
 #' @keywords AutoSpearman
-#' @examples 
+#' @examples
 #' Data = loadDefectDataset('groovy-1_5_7','jira')
 #' AutoSpearman(dataset = Data$data, metrics = Data$indep)
 #' @export
@@ -22,25 +22,14 @@ AutoSpearman <-
            spearman.threshold = 0.7,
            vif.threshold = 5,
            verbose = F) {
+    # Remove constant metrics and categorical metrics
+    metrics <- remove.constant.categorical(dataset, metrics)
 
-    # Check constant metrics
-    constant <- apply(dataset[, metrics], 2, function(x) max(x) == min(x))
-    constant <- names(constant[constant == TRUE])
-    # Remove constant metrics
-    if(length(constant) > 0){
-      metrics <- metrics[!metrics %in% constant]
-    }
 
-    # Check categorical metrics
-    category <- sapply(dataset[, metrics], class)
-    category <- names(category[category=="character"])
-    # Remove categorical metrics from Spearman Analysis
-    if(length(category) > 0){
-      metrics <- metrics[!metrics %in% category]
-    }
+    spearman.metrics <-
+      get.automated.spearman(dataset, metrics, spearman.threshold, verbose)
+    AutoSpearman.metrics <-
+      stepwise.vif(dataset, spearman.metrics, vif.threshold, verbose)
 
-    spearman.metrics <- get.automated.spearman(dataset, metrics, spearman.threshold, verbose)
-    AutoSpearman.metrics <- stepwise.vif(dataset, spearman.metrics, vif.threshold, verbose)
-
-    return(c(AutoSpearman.metrics,category))
+    return(AutoSpearman.metrics)
   }

R/get.vc.R

@@ -9,6 +9,10 @@
 #' @keywords VarClus
 
 get.vc <- function(dataset, metrics, similarity = 'spearman', varclus.threshold = 0.7){
+
+  # remove constant metrics and categorical metrics
+  metrics <- remove.constant.categorical(dataset, metrics)
+
   f <- as.formula(paste("~", paste(metrics, collapse = " + ")))
   vc <-
     Hmisc::varclus(f,

R/get.vc.correlation.groups.R

@@ -0,0 +1,30 @@
+#' Get correlation groups according to VarClus based on the absolute Spearman correlation coefficients between metrics
+#'
+#' This function makes life simple by providing a VarClus.
+#' @param dataset  a data frame for data
+#' @param metrics  a vector of characters or a vector of characters for independent variables
+#' @param similarity a character for similarity measures (e.g., Spearman rank correlation), default = spearman
+#' @param varclus.threshold a numeric for correlation coefficient threshold value
+#' @importFrom Hmisc varclus
+#' @keywords VarClus
+
+get.vc.correlation.groups <- function(dataset, metrics, similarity = 'spearman', varclus.threshold = 0.7){
+
+  # Remove constant metrics and categorical metrics
+  metrics <- remove.constant.categorical(dataset, metrics)
+
+  f <- as.formula(paste("~", paste(metrics, collapse = " + ")))
+  vc <-
+    Hmisc::varclus(f,
+                   similarity = similarity,
+                   data = dataset[, metrics],
+                   trans = "abs")
+
+  var.clusters <-
+    cutree(vc$hclust, h = (1 - varclus.threshold))
+  melted.data <- melt(var.clusters)
+  varclus.correlation.groups <- data.frame(metrics = row.names(melted.data), rank = var.clusters)
+  row.names(varclus.correlation.groups) <- NULL
+
+  return(varclus.correlation.groups)
+}

R/remove.constant.categorical.R

@@ -0,0 +1,32 @@
+#' Remove constant metrics and categorical metrics
+#' 
+#' @param dataset  a data frame for data
+#' @param metrics  a characters or a vector of characters for independent variables
+#' @keywords constant categorical
+#' @examples
+#' Data = loadDefectDataset('groovy-1_5_7','jira')
+#' remove.constant.categorical(dataset = Data$data, metrics = Data$indep)
+#' @export
+remove.constant.categorical <-
+  function(dataset,
+           metrics) {
+    # Check constant metrics
+    constant <-
+      apply(dataset[, metrics], 2, function(x)
+        max(x) == min(x))
+    constant <- names(constant[constant == TRUE])
+    # Remove constant metrics
+    if (length(constant) > 0) {
+      metrics <- metrics[!metrics %in% constant]
+    }
+
+    # Check categorical metrics
+    category <- sapply(dataset[, metrics], class)
+    category <- names(category[category == "character"])
+    # Remove categorical metrics from Spearman Analysis
+    if (length(category) > 0) {
+      metrics <- metrics[!metrics %in% category]
+    }
+
+    return(metrics)
+  }