modelEDA: Exploratory data analysis on a modeling dataset

In dnegrey/miscTools: R package with miscellaneous tools

Description

modelEDA performs standard exploratory data analysis on a modeling dataset, including variable clustering, weight of evidence (binary y), numeric R-squared (continuous y) and univariate summaries/graphs. Parameters prefaced with a function dot (i.e. bin.) are only applicable to the use of that function.

Usage

modelEDA(x, yname, ytype, bin.numBins = 10, bin.equalBinSize = FALSE,
  bin.minPct = 0, bin.maxPct = 100, nrsq.na.rm = FALSE,
  variable_cluster.n = max(floor((length(x) - 1)/10), 2),
  variable_cluster.na.rm = TRUE, univariateSummary.FUN = mean,
  univariateGraph.yLabel = yname, univariateGraph.yType = ifelse(ytype == 1,
  "pct", "dlr"), univariateGraph.yDigits = ifelse(ytype == 1, 1, 0),
  univariateGraph.yRangeMode = "tozero",
  univariateGraph.barColor = "#BDDFF7",
  univariateGraph.lineColor = "#000000")

Arguments

x	data frame; a modeling dataset
yname	character string; dependent variable column name
ytype	integer value of 1 (binary y) or 2 (continuous y)
bin.numBins	integer value >= 2; number of desired bins
bin.equalBinSize	logical value; return equally sized (TRUE) or equally spaced (FALSE) bins
bin.minPct	integer between 0 and 100 specifying a percentile to force as the max endpoint for the low (first) bin
bin.maxPct	integer between 0 and 100 specifying a percentile to force as the min endpoint for the high (last) bin (must be > minPct)
nrsq.na.rm	logical value indicating whether missing values of x (and their corresponding y values) should be removed
variable_cluster.n	integer value >= 2; number of desired clusters
variable_cluster.na.rm	logical value; should records with missing values be removed?
univariateSummary.FUN	function to be applied to y
univariateGraph.yLabel	character string; y variable label
univariateGraph.yType	character string; y variable format type; valid values are "int", "dlr" and "pct"
univariateGraph.yDigits	non-negative integer value indicating the number of decimal places to show for values of the y variable
univariateGraph.yRangeMode	character string; "tozero" (y-axis starts at 0) or "auto" (y-axis extremes determined by data)
univariateGraph.barColor	character string; fill color for bars (valid color)
univariateGraph.lineColor	character string; line color (valid color)

Value

A named list with class mt_modelEDA containing the following objects:

• dataSummaryDT: datatable() object summarizing data

• y.relativeHistogram: relativeHistogram() object for y

• variable_cluster: data frame with variable_cluster() results

• woe (binary y): named list of woe() tables

• infoValue (binary y): named list of infoValue() values

• woeDT (binary y): named list of woeDT() objects

• nrsq (continuous y): named list of nrsq() values

• variable_cluster_plus: data frame with variable_cluster_plus() results

• clusterDT: named list of clusterDT() objects

• univariateSummary: named list of univariateSummary() tables

• univariateSummaryDT: named list of univariateSummaryDT() objects

• univariateGraph: named list of univariateGraph() objects

See Also

woe, nrsq, variable_cluster, univariateGraph

Examples

# binary y 
x <- modelEDA(mtcars, "vs", 1)
names(x)
x$woeDT

# continuous y
x <- modelEDA(mtcars, "mpg", 2)
names(x)
x$clusterDT

dnegrey/miscTools documentation built on May 3, 2019, 2:57 p.m.