Table 1.

Properties and Limitations of the Proposed Graphical EDA Methods

	Methods	Properties	Limitations
One-dimensional  plots	ECDF	Nonparametric estimate of the cumulative  distribution. All estimates are on the  same scale Reveals differences in  distributions. Easy to find the location of  the quantiles	Not good for visualizing the shape of the  distributions. Not easy to adjust for other  variables
	Histogram	Reveals most frequent values. Good for  visualizing: the location of the  distributions-multimodality and  asymmetry	Depends on user defined bandwidth. Not  good to directly assess center and spread
	Boxplot	Substantial reduction of distribution. Good  for visualizing: the relative location of  the distributions–asymmetry. Samples  are compared on a vertically aligned  scale	Relevant for unimodal distributions. Not  good for visualizing the shape of the  distributions. The x-axis is typically  arbitrary
Two-dimensional  plots	Scatterplot	Statistical summary of multidimensional  data. Detection of outliers and their  relationship (e.g., plate effect)	Low throughput visualization. Comparison  of many data sources is difficult (require  multiple plots). Relevant for bivariate  distributions (can be extended).
	2D view  (contour plot)	Deals with large number of data points.  Detection of spatial variation and  association	Two-dimensional. Difficult to compare  many different views