Boxplots

Box plot PROC BOXPLOT, PROC GPLOT, or PROC UNIVARIATE... 

Note that for box and survival plots, PROC GPLOT is listed as an alternative. Although PROC BOXPLOT and PROC LIFETEST produce excellent graphs by themselves, sometimes it is necessary to make modifications to the output in a way that these procedures cannot handle directly. When modifications are needed, PROC GPLOT is an excellent choice. Also note that PROC REG and PROC UNIVARIATE are listed as options for scatter plots and box plots, respectively, as they can be useful in producing lower-resolution graphics for statistical appendices. 

There are other ways to create box plots in SAS. If you want a quick box plot, you can use PROC UNIVARIATE with the PLOT option. A more elegant option is to use PROC BOXPLOT, which was introduced in SAS 8. PROC BOXPLOT is explained in detail in the SAS/STAT User s Guide. The following is a sample of the same box plot as in the previous example, but created with PROC BOXPLOT. 

Here is the SAS program that creates this PROC BOXPLOT box plot. The initial input DATA step is omitted because it remains the same as in Programs 6.4 and 6.5. 

Program 6.6 Creating a Box Plot with Means Using PROC BOXPLOT... 

SORT DATA FOR PROC BOXPLOT. proc sort... 

DEFINE GRAPHICS OPTIONS SET DEVICE DESTINATION TO MS OFFICE CGM FILE, REPLACE ANY EXISTING CGM FILE, RESET ANY SYMBOL DEFINITIONS, AND DEFINE DEFAULT FONT TYPE. filename fileref C whisker proc boxplot.cgm goptions... 

CREATE BOX PLOT. VISIT IS ON THE X AXIS, SEIZURES ARE ON THE Y AXIS, AND THE VALUES ARE PLOTTED BY TREATMENT. proc boxplot... 

Note that PROC BOXPLOT uses SAS/GRAPH statements such as GOPTIONS and SYMBOL statements. There is a vast array of display options in the PROC BOXPLOT PLOT statement not used here. You may want to investigate those options if you decide to enhance the box plot further. 

The most important methods of explorative data analysis concern the study of the distribution of the data and the recognition of outliers by boxplots (Fig. 8.18), histograms (Fig. 8.19), scatterplot matrices (Fig. 8.20), and various schematic plots. 

Fig. 8.18. Representation of some trace elements in wines in form of boxplots, constructed as follows box lower quartile, median , and upper quartile whiskers minima and maxima within box 1.5 of the quartiles difference outliers 0 outside of box 1.5 of quartiles difference (according to Danzer et al. [2001])...

Fig. 2 Boxplots of the concentrations of individual and total pesticides and frequency of detection (expressed in % in parentheses) in the water samples collected from each of the six sampling sites. HDAD, HDCD, HEOD, and HEIMD draining channels HDM and HEM bays DIA Deisopro-pylatrazine DEA desethylatrazine...

Fig. 32.2 Responses of pre-term infants to the odours of different milk stimuli at 36 weeks PCA. Boxplots that do not share the same super-script letter differed significantly...

FIGURE 1.6 Boxplot of the CaO data with definitions and outlier boundaries (left) and parallel boxplots of the CaO data for four individual groups of glass vessels (right). 

Some of the above plots can be combined in one graphical display, like onedimensional scatter plot, histogram, probability density plot, and boxplot. Figure 1.7 shows this so-called edaplot (exploratory data analysis plot) (Reimann et al. 2008). It provides deeper insight into the univariate data distribution The single groups are... 

FIGURE 1.7 EDAPLOT of data combines one-dimensional scatter plot, histogram, probability density trace, and boxplot. Data used are CaO concentrations (%) of 180 archaeological glass vessels. 

The pdf-file FileName.pdf is generated with the size 6 in. times 4 in. containing a boxplot of x. 

Figure 4. Boxplot summaries of the elemental data. The upper plot shows the plaza data (n = 182) and the lower plot shows the patio data (n = 142).

Figure 8 Boxplots of normalized intensity values across arrays, boxplots facilitate comparison of sample distributions and dynamic range across microarrays. If the microarrays follow different data distributions increased false positive and negative rates may be demonstrated. The depicted signal is the log2 of the normalized feature signal intensity.

Figure 3. Trend in annual median values of sulfate concentration in bulk deposition samples at Hinckley, NY (crosses). Variability is also indicated in the boxplot by the standard deviations (-) and the range (vertical bars) and outliers (0, )(Reproduced with permission from Ref. 4. Copyright 1984 APCA).

Void content for any pair of load-angle data, produces one value, therefore, normality and homogeneity of all 5 replicates should be verified to report a mean value. Boxplot, Anderson-... 

---