• Print |
• E-mail  |
• Bookmark

If the PLOT option is specified on the PROC UNIVARIATE statement,
BY-group processing is used, and if one or more BY groups have either
constant or all missing values for the analysis variable then the
side-by-side box plots will be incorrect.  When only the first BY group
has constant or all missing values, the box plots themselves are
correct but the labels will all be shifted over one position to the left
and the last BY group label is used twice. That is, the value of the
second BY variable(s) appears under the first box plot, the value of the
third BY variable(s) appears under the second box plot, etc. The last
two boxes are both labeled with the value of the last BY variable(s). If
any BY group beyond the first contains constant or all missing values
for the analysis variable, then one or more of the b ox plots will be
incorrect in addition to the labels being incorrect. The problem is
always identifiable by that fact that two of the box plots will have the
same label.

To circumvent this problem, omit the side-by-side box plots with the
statement

     ODS LISTING EXCLUDE SSPLOTS;

and instead use either the BOXCHART statement in PROC SHEWHART or PROC
GPLOT with the option INTERPOL=BOX on a SYMBOL statement or PROC
BOXPLOT. Note that in the default plots produced by PROC SHEWHART, PROC
GPLOT and PROC BOXPLOT, the whiskers are drawn to the extreme values in
each BY group. Refer to the Online Documentation for options on
alternate specifications.

Operating System and Release Information