1. Literate Programming in SAS
Bruce Schaalje
Gauss June 2018

2. technical definition:
Literate Programming
technical definition:
programming paradigm in which program logic is given in a natural language, interspersed with source code.
in practice:
integration of report-creation code and statistical code to "create dynamic reports, which can be updated automatically if data or analysis change.1”
1 Leisch, 2002, “SWEAVE, Part 1”

3. Rajaram et al., 2017, “Automated Generation of Clinical Study Reports”

4. Rajaram et al., 2017, “Automated Generation of Clinical Study Reports”

5. Example Code *******************************************************;
****** automated analysis and report generation *******;
** specify report-specific names and parameter values *;
%let project=Report 123;
%let crop=Soybean;
%let material=OK-999;
%let matcode=1;
%let reportname=literate example;
%let control=OK-888;
%let controlcode=1;
%let nmat=25;
%let nmatword=Twenty five;
%let nref=23;
%let nrep=2;
%let year=1958;
%let location=Iowa;
%let statistician=Gertrude M. Cox, Senior Statistician;
********************************************************;
******** "here is the line" ****************************;
******* templates and formats **************************;
/* This template creates rtf tables with Times New Roman
12 pt font, and margins of 1 inch on all four sides */
PROC TEMPLATE;
DEFINE STYLE STYLES.required/STORE=WORK.RPTTABLE;
PARENT = STYLES.JOURNAL;
STYLE FONTS From FONTS/
'SASTitleFont' = ("Times",12pt)
'TitleFont2' = ("Times",12pt)
'TitleFont' = ("Times",12pt)
'StrongFont' = ("Times",12pt)
'EmphasisFont' = ("Times",12pt)
'FixedEmphasisFont' = ("Times",12pt)
'FixedStrongFont' = ("Times",12pt)
'FixedHeadingFont' = ("Times",12pt)
'FixedFont' = ("Times",12pt)
'headingEmphasisFont' = ("Times",12pt)
'headingFont' = ("Times",12pt)
'docFont' = ("Times",12pt);
REPLACE BODY FROM DOCUMENT /
LEFTMARGIN = 1.3in
RIGHTMARGIN = 1.0in
TOPMARGIN = 1.0in
BOTTOMMARGIN = 0.5in;
STYLE CONTINUED FROM CONTINUED /
PRETEXT=" "
FONT=("ARIAL", 1pt) WIDTH=1%;
STYLE PARSKIP FROM PARSKIP /
FONT=("ARIAL", 1pt);
END;
RUN;
** create paragraph formats and ods escape character;
%let textfont1=%str(^S={font_size=12pt
font_face='Times Roman' just=c});
%let textfont2=%str(^S={font_size=12pt
font_face='Times Roman' just=l font_weight=bold});
%let textfont3=%str(^S={font_size=12pt
font_face='Times Roman' just=j});
%let textfont4=%str(^S={font_size=12pt
font_face='Times Roman' just=c font_weight=bold});
%let textfont5=%str(^S={font_size=12pt
font_face='Times Roman' just=l});
ods escapechar='^';
options nodate nonumber;

6. Example Code
*********************************************************;
*********** start ods rft document **********************;
ods rtf style=required file="&reportname statistics
report.rtf" NOTOC_DATA;
footnote1 h=10pt j=l "{ABC Consulting}"
j=c "{Page }^{pageof}"
j=r "&project";
**********************************************************;
******* Specify titles, subtitles, and paragraphs. *******;
**** Note: Line breaks can be inserted anywhere. *********;
**** However, the next line must start with a space. *****;
************** report title and introduction *************;
ods rtf text = "&textfont1
{Statistical Report}";
ods rtf text=" ";
ods rtf text="&textfont4
{Evaluation of &crop &material in &year &location
Field Trial}";
ods rtf text="&textfont1
{Study Number: &project}";
ods rtf text="&textfont2
{Purpose}";
ods rtf text=" ";
ods rtf text="&textfont3
{The purpose of this analysis was to test whether the
average yield of the test material &crop &material
was different than the average yield of the conventional
control &crop &control or the average yields of various
reference materials.}";
********************************************************;
************* Input and prepare data *******************;
ods rtf text="&textfont2
{Data Description}";
{&nmatword materials were planted in &Location, with one
test material (&material), one conventional control
material (&control), and &nref reference materials. The
experimental design was a 5x5 square lattice with &nrep
complete replicates and 5 incomplete blocks per
replicate. Yield data (bu/acre) were collected at
harvest. Yield data were input directly into a SAS (SAS 2012)
data set for statistical analysis.}";
data new;
input rep block variety yield
cards;
run;

7. Example Code *******************************************************;
********** Statistical Methods and Results ************;
ods rtf text=" ";
ods rtf text="&textfont2
{Statistical Methods and Results}";
ods rtf text="&textfont3
{(1) Summary Statistics}";
{PROC MEANS in SAS was used to calculate means and
standard deviations of yield for all experimental
materials. The results are presented in Table 1.
the relationship between the means and the
standard deviations of the materials is displayed
in Figure 1.}";
********************************************************;
************** Means and Standard Deviations ***********;
proc sort data=new;
by variety rep;
run;
proc means data=new noprint;
by variety;
var yield;
output out=m1 mean=mean std=std;
*********************************************************;
****************** Statistical Analysis *****************;
ods rtf text=" ";
ods rtf text="&textfont3
{(2) Statistical Analysis}";
{The following statistical model was used to analyse
the data:}";
y^{sub ijk} = ^S={font_face=symbol
font_size=12pt}m + ^S={font_face='Times Roman'
font_size=12pt}r^{sub i} + t^{sub j} + b^{sub k} +
e^{sub ijk} (1) ^2n where
y^{sub ijk} = yield (bu/acre); ^S={font_face=symbol
font_size=12pt}m ^S={font_face='Times Roman'
font_size=12pt} = overall mean; r^{sub i} = random
replicate effect; t^{sub j} = fixed material effect;
b^{sub ik} = random block effect; and e^{sub ijk} =
residual effect.";
{PROC MIXED was used to fit model (1) to the data.
Comparisons between the test and other (control and
reference) materials were defined within the SAS
code, and tested using t-tests. The results of the
comparisons are displayed in Table 2.}";

8. Example Code
*********************************************************;
****************** statistical analysis *****************;
ods exclude all;
ods graphics on;
proc mixed data=new;
class rep block variety yield;
model yield = variety / ddfm=kr;
random rep block(rep);
lsmeans variety / pdiff=all;
ods output diffs=indcomps tests3=test;
run;
ods exclude none;
/* Summarizing significance */
data test;
set test;
if Probf gt 0.05 then sig='not';
else sig ='';
call symput('sig',sig);
data inddiffs;
set indcomps;
if variety=&matcode;
if Probt lt 0.05 then sig='*';
else sig =' ';
if Probt lt 0.05 then signif=1;
else signif=0;
proc means noprint sum data=inddiffs;
var signif;
output out=s1 sum=sum n=n;
data s1;
set s1;
call symput('sum',sum);
call symput('n',n);
ods rtf text=" ";
ods rtf text="&textfont2{
Summary of statistical analysis results}";
ods rtf text="&textfont3
{Comparisons between the test and other materials
revealed %left(&sum) significant differences out of
%left(&n) tests (Table 2).}";
*********************************************************;
****************** insert references ********************;
data null;
x=" ";
run;
ods startpage=no;
* used to start a new page the first time;
proc report data=null nowd noheader
style=[bordercolor=white];
column x;
ods startpage=now;
ods rtf text="&textfont2
{Reference}";
{SAS (2012). Software Release 9.4 (TS1M4).
Cary, North Carolina, Copyright by
SAS Institute, Inc.}";

9. Example Code
*********************************************************;
*************** insert tables with headings ************;
ods startpage=now;
ods rtf text="^S={font_size=12pt font_face='Times Roman'
font_weight=bold}
{Table 1. }
^S={font_size=12pt font_face='Times Roman'}{Summary
Statistics for Yield Data}";
ods rtf text=" ";
proc print noobs data=m1 label;
format mean std 6.1;
var variety mean std;
label variety='Soybean Variety' mean='Mean Yield'
std='Standard Deviation of Yield';
run;
{Table 2. }
^S={font_size=12pt font_face='Times Roman'}{Statistical
Comparisons of Varieties}";
proc print noobs data=inddiffs label;
var _variety estimate stderr probt sig;
label _variety='Soybean Variety'
estimate='Difference from Control'
stderr='Standard Error' probt='P-value'
sig='Significant';
*******************************************************;
************** insert figure with caption ***********;
ods startpage=now;
ods rtf text="^S={font_size=12pt font_face='Times Roman'
font_weight=bold} { Figure 1. }
^S={font_size=12pt font_face='Times Roman'}
{Mean Versus Standard Deviation for Soybean Yields.}";
ods rtf text=" ";
ODS graphics on / imagefmt=png noborder;
footnote;
proc sgplot data=m1;
scatter y=mean x=std;
reg y=mean x=std/nomarkers;
run;
footnote1 h=10pt j=l "{ABC Consulting}"
j=c "{Page }^{pageof}"
j=r "&project";
ods startpage=no;

10. Example Code
**********************************************************;
**************** insert signature page *******************;
data signature;
x=" ";
y=" ";output;
x=" ";
y=" ";output;
x="&statistician";y="Date";output;
x="Statistics Team";y=" ";output;
x="ABC Consulting";y=" ";output;
run;
ods startpage=now;
ods rtf text="&textfont2
{SUBMITTED BY:}";
proc report data=signature nowd noheader
style=[bordercolor=white] style=[paddingright=10]
style=[cellpadding=0];
column x y;
ods rtf close;

11. Example Output

12. Example Output

13. Example Output

14. Advantages of Literate Programming using SAS RTF
Fast
SAS calculates statistics, creates report, creates tables, inserts tables, numbers pages, inserts footnotes, formats report
Easy to revise (one step)
Easy to create multiple similar reports
Accurate
No mistakes in inserting tables, no mistakes in page numbers and footnotes
Universal name changes
SAS can count and insert #tests, etc.
Standardized format details can be enforced in all reports (justification, etc.)

15. Advantages of Literate Programming using SAS RTF
GLP friendly
All calculations and revisions documented
All other changes documented
Grammar friendly
Document can be word-smithed and locked
Protocol friendly
Accurate details of protocol can be specified using %let statements
Flexible
Can be adapted to various degrees of automation

16. Disadvantages of Literate Programming using SAS RTF
Initial setup time
“above the line” details have to be specified
text must be initially inserted into SAS
formatting details must be added
Learning curve
for ODS text, formatting, escape characters
pretty good guidance in SUGI reports
SAS documentation not-so-good

17. Disadvantages of Literate Programming using SAS RTF
Some quirky things about SAS ODS RTF
but all can be solved or worked around

18. Useful References To ODS RTF and Beyond SAS – tricks and tips
SAS Guide to Report Writing
Automated Generation of Reports
Reproducible RTF memos

19. Thank You