Example Program and Statement Details

The following graph was generated by the Example Program:

proc template;
  define statgraph boxplotparm1;
    begingraph;
      entrytitle "City Mileage for Vehicle Types";
      layout overlay;
       boxplotparm y=value x=x stat=stat /
         datalabel=datalabel spread=true ;     
      endlayout;
    endgraph;
  end;
run;
 
proc sgrender data=boxdata template=boxplotparm1;
run;

The following input data generated the box for Sedan in the graph. See Generalized Macro for BOXPLOTPARM Data to see the code for creating all of the data.

STAT      X          VALUE    DATALABEL
 ...      
N         Sedan        262
MEAN      Sedan    21.0840
MEDIAN    Sedan         20 
Q1        Sedan         18 
Q3        Sedan         24 
STD       Sedan     4.2346
OUTLIER   Sedan         36    Honda
OUTLIER   Sedan         35    Toyota
OUTLIER   Sedan         35    Toyota
OUTLIER   Sedan         38    Volkswagen
MIN       Sedan         12 
MAX       Sedan         33 
 ...

The BOXPLOTPARM statement requires pre-computed input data. One reason to choose this statement over the BOXPLOT statement is that you can control the computational technique used to define the box plot components: the mean, quartiles, location of fences, outlier definition, and so on. See Generalized Macro for BOXPLOTPARM Data for examples of such computations using PROC SUMMARY and DATA steps.

The BOXPLOTPARM statement displays a single box if given just Y and a STAT argument. It displays multiple boxes if given both Y and X and a STAT argument and X has more than one unique value.

Two basic box plot representations can be drawn with the BOXPLOTPARM statement: a schematic (Tukey) box plot and a skeletal box plot. See the EXTREME= option for details.

The following figure illustrates the box plot elements:

As shown in the figure, the bottom and top edges of the box are located at the 1st quartile (25th percentile) and 3rd quartile (75th percentile) of the sample. Within the box you can display the median (50th percentile) as a line and the mean as a marker (see the DISPLAY= option).

You can also display markers and data labels for outliers. Outliers are observations that are more extreme than the upper and lower fences ( ). Outliers that are beyond upper and lower far fences ( ) are called FAR OUTLIERS and can also be identified and labeled. From a graphical perspective, the location of fences along the axis are known, but there is no line or marker that displays a fence. (See DISPLAY=, LABELFAR=, and DATALABEL= options).

Finally, you can control the range represented by the whiskers. By default, the whiskers are drawn from the upper edge of the box to the MAX value, and from the lower edge of the box to the MIN value (see EXTREME= option).

At a minimum, valid data for the BOXPLOTPARM statement must provide a numeric column (Y=) that contains calculated statistics for an analysis, and a string column (STAT=) that identifies each statistic. The Y column must contain nonmissing values for the Q1 (25th percentile) and Q3 (75th percentile) statistics. If Y values are missing or not supplied for other statistic values, then those statistics are not displayed in the plot, regardless of syntax requests to display them.

For example, a petroleum company uses a turbine to heat water into steam that is pumped into the ground to make oil more viscous and easier to extract. This process occurs 20 times daily, and the amount of power (in kilowatts) used to heat the water to the desired temperature is recorded. The following data show the statistics that are calculated for one day of this process:

To plot the data from the preceding table, the following BOXPLOTPARM statement uses the Y= and STAT= arguments to generate a single box plot for the recorded statistics:

BOXPLOTPARM Y=PowerOutputs STAT=Statistic; 

If the data contain statistics for multiple days of the process, a third column in the data must be present to identify the days that the statistics were recorded. For example, the following data show the statistics that are calculated for two days of this process:

To plot the data from the preceding table, the BOXPLOTPARM statement needs the Y=, STAT=, and X= arguments to generate a separate box plot for each day that the statistics were recorded:

   BOXPLOTPARM Y=PowerOutputs STAT=Statistic X=Day; 

See Generalized Macro for BOXPLOTPARM Data for a more complete example of providing input data for BOXPLOTPARM.