### SAS FAQ How do I make a histogram with percentage on top of each bar?

NOTE: As of SAS 9.2, the histogram statement in proc univariate will now by default direct graphs to ODS graphics rather than "tradtional graphics". Many old options, such as cfill=, which was used to change the color of the histogram bars, are ignored by ODS graphics and have been replaced by style options that can be set in proc template. Previous versions of this page used traditional graphics, the code for which we have preserved at the bottom of the page. The page has been updated for SAS 9.3 with creation of histograms with percentages on the bars using ODS graphics.

Consider this simple data file with a variable called mynum. We are going to create a histogram of mynum with percentages on top of each bar.

data test;
input mynum @@;
cards;
2.3 -.6
3.3 3.5
2.4 5.6
7.8 2.4
2.8 4.5
6.3 1.2
0.5  .8
.9  1.2
1.4 1.5
2.3 2.5
2.7 3.5
3.1 4.6
5.5 5.8
5.3 7.6
7.3 7.8
;
run;

Let's say the midpoints that we are going to use for our histogram are -.5, 1.5, ... 7.5, since the range of the variable is between -1 to 8. The width of each bin is 1. Below in a data step, a variable called ind is created as a group variable representing each bin (bar). Then we use proc freq with ods output option to output the frequencies and percentages to a dataset called temp3

data temp2;
set test;
do i = -.5 to 7.5 by 1;
if  i-.5 <= mynum < i+.5 then ind=i;
end;
run;
proc freq data=temp2;
tables ind ;
ods output OneWayFreqs=temp3;
run;

Let's print out the ods output data set temp3 from proc freq.

proc print data=temp3;
run;

Cum         Cum
Obs    Table    F_ind     ind    Frequency    Percent    Frequency    Percent

1      ind     -0.5     -0.5           1       3.33            1       3.33
2      ind      0.5      0.5           3      10.00            4      13.33
3      ind      1.5      1.5           4      13.33            8      26.67
4      ind      2.5      2.5           7      23.33           15      50.00
5      ind      3.5      3.5           4      13.33           19      63.33
6      ind      4.5      4.5           2       6.67           21      70.00
7      ind      5.5      5.5           4      13.33           25      83.33
8      ind      6.5      6.5           1       3.33           26      86.67
9      ind      7.5      7.5           4      13.33           30     100.00

We are now ready to create our histogram with percentages on top of each bar. We request these precentages with the option barlabel=percent in the histogram statement. In proc template, we define a new style newstyle that inherits all of the specification of the SAS HTML default style htmlblue, but we change the color of the bars and the color of the text used to print the percentages. We then ask that SAS use this new style with the ods html style= statement.


proc template ;
define style Styles.newstyle;
parent = Styles.htmlblue;
Color = pink;
style GraphDataText /
Color = blue;
end;
run;

ods html style=Styles.newstyle;

proc univariate data=temp2 ;
histogram mynum /barlabel=percent midpoints=-.5 to 7.5 by 1;
run;


NOTE: The following text and code, created before SAS 9.2, are from an older version of this page, which will create the same histogram produced above, but here in SAS "traditional graphics". To create this histogram with traditional graphics, so that all the options listed here will work, the command ods graphics off; must be issued.

Consider this simple data file with a variable called mynum. We are going to create a histogram of mynum with percentages on top of each bar.

data test;
input mynum @@;
cards;
2.3 -.6
3.3 3.5
2.4 5.6
7.8 2.4
2.8 4.5
6.3 1.2
0.5  .8
.9  1.2
1.4 1.5
2.3 2.5
2.7 3.5
3.1 4.6
5.5 5.8
5.3 7.6
7.3 7.8
;
run;

Let's say the midpoints that we are going to use for our histogram are -.5, 1.5, ... 7.5, since the range of the variable is between -1 to 8. The width of each bin is 1. Below in a data step, a variable called ind is created as a group variable representing each bin (bar). Then we use proc freq with ods output option to output the frequencies and percentages to a dataset called temp3

data temp2;
set test;
do i = -.5 to 7.5 by 1;
if  i-.5 <= mynum < i+.5 then ind=i;
end;
run;
proc freq data=temp2;
tables ind ;
ods output OneWayFreqs=temp3;
run;

Let's print out the ods output data set temp3 from proc freq.

proc print data=temp3;
run;

Cum         Cum
Obs    Table    F_ind     ind    Frequency    Percent    Frequency    Percent

1      ind     -0.5     -0.5           1       3.33            1       3.33
2      ind      0.5      0.5           3      10.00            4      13.33
3      ind      1.5      1.5           4      13.33            8      26.67
4      ind      2.5      2.5           7      23.33           15      50.00
5      ind      3.5      3.5           4      13.33           19      63.33
6      ind      4.5      4.5           2       6.67           21      70.00
7      ind      5.5      5.5           4      13.33           25      83.33
8      ind      6.5      6.5           1       3.33           26      86.67
9      ind      7.5      7.5           4      13.33           30     100.00

We can now use this data set to create an annotate data set and use it with proc univariate to create the histogram with percentage on top of each bar.

data anno;
set temp3;
length function color text $8; function = 'label'; color = 'blue'; size = 1; xsys = '2'; ysys = '2'; when = 'a'; x=ind; /*the x-coordinate for the text*/ y=percent+.5; /*the y-coordinate*/ text=left(put(percent, 4.2)); run; ods graphics off; proc univariate data=temp2 noprint; histogram mynum /anno=anno cfill=pink midpoints=-.5 to 7.5 by 1; run; Using the same idea, we can also get a histogram with cumulative percentage on top of each bar. This time we lower the Y-axis position so the percentage is inside of the bar.  data anno; set temp3; length function color text$8;

function = 'label';
color    = 'blue';
size     =  1;
xsys     = '2';
ysys     = '2';
when     = 'a';
x=ind;
y=percent-.5;
text=left(put(cumpercent, 4.2));
run;

ods graphics off;

proc univariate data=temp2 noprint;
histogram mynum /anno=anno cfill=pink midpoints=-.5 to 7.5 by 1;
run;

The content of this web site should not be construed as an endorsement of any particular web site, book, or software product by the University of California.