NiceHist Histogram with Density

NiceHist is a tool to examine a metric variable in data inspection, cleaning and description. A histogram with density curve is produced. ScaleTransformations: ID, POWER_Z, LOG10, LOGIT are supported. Location and scale parameter as well as information on valid values can be displayed as legend. Outliers can be identified and filtered out using a +- factor* mad filter.

NiceHist(METRIC, NCLASS = 40, ADJ = 1, SCALE = "ID", MADFILTER = 0,
  MAIN = "Histogram with Density", XLAB = NULL, YLAB = "Density",
  XLIM = NULL, NPRETTY = 6, RUG = T, HISTCOL = 7, DENSCOL = 4,
  LEGEND = T, DIGITS = 3, CUTLINE = NULL, ...)

Arguments

METRIC	metric variable
NCLASS	= 40 defines the number of classes (=bins) in the histogram
ADJ	= 1controls the smoothing of the density curve relative to default=1
SCALE	="ID" chooses a ScaleTransformation ID, POWER_Z, LOG10, LOGIT for X: use optional parameter Z to specify the power to use with POWER_Z. use optional parameter PERCENT = T for percent instead of proportions on LOGIT scale.
MADFILTER	=0, if >0 (typically 3,4, or 5) filters out Outliers beyond +- MADFILTER* mad
MAIN	= "Histogram with Density" specifies the title of plot
XLAB	specifies the x-axis-Label, if NULL the variable name of x is extracted from the call using PruneVarName
YLAB	="Density" specifies the y-axis-Label
XLIM	= NULL sets the x-axis range to be plotted manually - generally not necessary
NPRETTY	=6 influences approximate number of labels when using ID or POWER_Z (cf. base pretty)
RUG	=T logical whether to plot individual values above x-axis
HISTCOL	defines the colour of the histogram, default=7 (yellow)
DENSCOL	defines the colour of the density curve, default=4 (blue). DENSCOL=NA suppresses the density curve,
LEGEND	= T show legend, which is smartly placed on the side where there is most space.
DIGITS	controls the numbers of digits shown for location/scale parameters in the LEGEND
CUTLINE	=NULL Draws vertical lines from cut values to the density curve
...	further parameters for scale transformations can be provided via ... Z specifies the power to use with POWER_Z when using POWER_Z PERCENT =F Use percent instead of proportions in LOGIT when using LOGIT

Value

data.frame Irregular with 4 variables: Rownr, and 2 logicals TransUndef and Outlier in order to identify irregular values.

See also

hist

Examples

if (FALSE) {
Metric<-c(rnorm(200),rnorm(100,3,1))
NiceHist(Metric)
NiceHist(Metric,RUG=F,ADJ=.7)
NiceHist(Metric,LEGEND = F)

# Demo LOG10 scale
NiceHist(exp(Metric),LEGEND = T,SCALE="ID",NCLASS=100)
NiceHist(exp(Metric),LEGEND = T,SCALE="ID",NCLASS=1000,XLIM=c(0,50))
NiceHist(exp(Metric),LEGEND = T,SCALE="LOG10")

# Demo POWER_Z scale
Metric<-rnorm(300,5,1)^3
NiceHist(Metric)
NiceHist(Metric,SCALE="POWER_Z",Z=1/3,ADJ=.8)

# Demo LOGIT scale
Metric<-runif(300)^3
NiceHist(Metric)
NiceHist(Metric,SCALE="LOGIT")
NiceHist(Metric,SCALE="LOGIT",LEFT=T)
NiceHist(100*Metric,SCALE="LOGIT",LEFT=T,PERCENT=T)
NiceHist(100*Metric,SCALE="LOGIT",LEFT=T,PERCENT=T,XLAB="Percent values")

# Dealing with outliers and missing values
Metric<-c(rnorm(200),rnorm(100,3,1),runif(10,5,100),rep(NA,5))
NiceHist(Metric)
Irregular<-NiceHist(Metric,MADFILTER=4)
Metric[Irregular$Outlier]
}