PhUSE 2017

Paper CT02

Efficient Coding Techniques In SAS®

Geetha Kesireddi, GCE Solutions Inc, Hyderabad, India

ABSTRACT:
SAS is most widely used language in Clinical industry for statistical analysis and its usage is improving drastically.
Regular habituated way of programming in SAS is usage of some of the old functions, regular ‘if then else’
statements. These are basic techniques which are widely used. However, as the industry is growing, programming
using these basic techniques is making the code humongous and difficult for maintenance. Continuous upgrade of
both hardware and software over time, is becoming very difficult with these basic methods/approach. One of the
solutions for this problem is improving the coding techniques by using new functions like IFC, IFN, CHOOSEC and
CHOOSEN. Which can be used in day to day programming, to make the code much easy for maintenance and to
drastically brings down the time to upgrade.

INTRODUCTION:
Almost every SAS programmer uses ‘If then else’ statements often to implement conditional logics in the program.
IFC, IFN, CHOOSEC and CHOOSEN functions are great alternatives to the regular ‘if then else’ statements. These
functions are very handy to use. These functions help in reducing the lines of code and helps in handling missing
values.

IFC, IFN functions are very useful for encoding, decoding and flagging values. These functions can be used in data
step and in proc SQL, making it more powerful.

This paper illustrates various scenarios where these functions can be used.

IFC FUNCTION:
The IFC function returns a character value based on whether an expression is true, false, or missing.

SYNTAX:

IFC(logical-expression, value-returned-when-true, value-returned-when-false <, value-returned-when-missing>)

EXAMPLE 1:

To derive treatment group variables (TRTA and TRTP). Based on the respective APERIOD values, TRTA and TRTP
values were assigned from TRT01A/TRT01P, TRT02A/TRT02P, TRT03A/TRT03P. Suppose if APERIOD is 1 then
TRTA/TRTP will be TRT01A/TRT01P etc.

SOLUTION 1: Using ‘if then else’ statements.

if APERIOD=1 then TRTA=TRT01A;

else if APERIOD=2 then TRTA=TRT02A;
else if APERIOD=3 then TRTA=TRT03A;

1
if APERIOD=1 then TRTP=TRT01P;
else if APERIOD=2 then TRTP=TRT02P;
else if APERIOD=3 then TRTP=TRT03P;

SOLUTION 2: Using ‘IFC’ function.

TRTA=ifc(APERIOD=1,TRT01A, ifc(APERIOD=2,TRT02A, ifc(APERIOD=3,TRT03A,'')));

TRTP=ifc(APERIOD=1,TRT01P, ifc(APERIOD=2,TRT02P, ifc(APERIOD=3,TRT03P,'')));

which yields the results:

EXAMPLE 2:

Conditional execution of statements using IFC. ENREF is created using AEENRF, AEENDTC, AEENDY variables
where in, the ‘U’ value in AEENRF should be changed to ‘UNKNOWN’ and if the AEENRF values are missing then in
that place AEENDTC and AENDY should be populated.

SOLUTION 1: Using ‘if then else’ statements.

data test;
set ae;
if AEENRF ne '' then do;
if AEENRF='U' then AEENRF='UNKNOWN';
ENREF=AEENRF; end;
else ENREF=cats(AEENDTC,'/',put(AENDY,best.));
run;

SOLUTION 2: Using 'IFC’ in proc SQL.

proc sql;
create table test
as select SUBJID,AETERM,AEENDTC,AAEENRF,AENDY,
ifc(AEENRF ^='',ifc(AEENRF='U','UNKNOWN',AEENRF),
ifc(AEENDTC ^='',cats(AEENDTC,'/',put(AENDY,best.)),'')) as ENREF
from ae;
quit;

which yields the results:

2
EXAMPLE 3:

For creation of flags like treatment completion flag (TRTCPLFL). If the subject’s decode value in disposition dataset is
‘COMPLETED’ in treatment epoch then treatment completion flag will be ‘Y’, else the reason for non-completion will
be specified in the treatment discontinuation (TRTDISC) column.

SOLUTION 1: Using ‘if then else’ statements.

if upcase(DSDECOD_TRT) eq 'COMPLETED' then TRTCPLFL='Y';

else TRTCPLFL='N';
if upcase(DSDECOD_TRT) eq 'COMPLETED' then TRTDISC='';
else TRTDISC = DSDECOD_TRT;

SOLUTION 2: Using ‘IFC’ function.

TRTCPLFL = ifc((upcase(DSDECOD_TRT) eq 'COMPLETED'), 'Y', 'N');

TRTDISC = ifc((upcase(DSDECOD_TRT) eq 'COMPLETED'), ' ', DSDECOD_TRT);

which yields the results:

IFN FUNCTION:
The IFN function returns a numeric value based on whether an expression is true, false, or missing.

SYNTAX: IFN(logical-expression, value-returned-when-true, value-returned-when-false<, value-returned-when-

missing>)

3
EXAMPLE 1:

To derive ATOXGRN and BTOXGRN (numeric variables of toxicity grades) in LAB dataset.

Sample data as follows:

SOLUTION 1:

• Using ‘if then else’ statements.

if ATOXGR ne '' and BTOXGR ne '' then do;

if ATOXGR='No grade' then ATOXGRN=0;

else ATOXGRN=input(scan(ATOXGR,2),best.);

if BTOXGR='No grade' then BTOXGRN=0;

else BTOXGRN=input(scan(BTOXGR,2),best.);
end;

SOLUTION 2:

• Using ‘IFN’ Function

if cmiss(ATOXGR,BTOXGR) eq 0 then do;

ATOXGRN = ifn(compress(ATOXGR,'No Grade') = '', 0,input(compress(ATOXGR,

'No Grade'),best.));

BTOXGRN = ifn(compress(BTOXGR,'No Grade') = '', 0,input(compress(BTOXGR,

'No Grade'),best.));

end;

which yields the results:

EXAMPLE 2:

To derive VISITNUM from VISIT variable.

4
Sample visit data as follows:

proc format;
value $vsnum
'RANDOMIZATION' = 100
'WEEK 1' = 101
'WEEK 2' = 102
'WEEK 4' = 103
'WEEK 8' = 104
'WEEK 12' = 105
'WEEK 13' = 106
'WEEK 14' = 107
'WEEK 15' = 108
'WEEK 16' = 109
'WEEK 20' = 110
;
run;

SOLUTION 1:

• Using ‘if then else’ statements.

data test;
set visit;
if VISIT = 'UNP_IVR' then VISITNUM = 999;
else VISITNUM = input(VISIT, $vsnum.);
run;

SOLUTION 2:

• Using ‘IFN’ Function

data test;
set visit;
VISITNUM = ifn((VISIT eq 'UNP_IVR'), 999, input(VISIT, $vsnum.));
run;

5
which yields the results:

CHOOSEC FUNCTION:
SYNTAX:

CHOOSEC (index-expression, selection-1 <,...selection-n>)

Index-expression value is to select from the arguments that follow, selection-1 to selection-n are the character values
from which to choose. If the first argument is negative, the function counts backwards from the list of arguments, and
returns that value.

Purpose is to select a character value from a list of character values (supplied as the 2 nd to nth arguments of the
function)

In a DATA step, if the CHOOSEC function returns a value to a variable that has not previously been assigned a
length, then that variable is given a length of 200 bytes.

EXAMPLE:

To assign APERIODC values based on APERIOD. If the APERIOD value is 1 then APERIODC will be ‘INDUCTION’,
if APERIOD value is 2 then APERIODC will be ‘MAINTENANCE’, if the APERIOD value is 3 then APERIODC will be
‘TREATMENT’.

SOLUTION 1:

• Using ‘if then else’ statements.

data test;
length APERIODC $20;
set vs;

6
 if APERIOD=1 then APERIODC='INDUCTION';
else if APERIOD=2 then APERIODC='MAINTENANCE';
else if APERIOD=3 then APERIODC='TREATMENT';
else APERIODC='';
run;

SOLUTION 2:

• Using ‘CHOOSEC’ function.

data test;
length APERIODC $20;
set vs;
APERIODC=choosec(APERIOD, 'INDUCTION', 'MAINTENANCE', 'TREATMENT', '');
run;

Hence based on the index expression value of aperiod, CHOOSEC function assigns the appropriate aperiodc values
which are listed.

Which yields the results:

CHOOSEN FUNCTION:
SYNTAX:

CHOOSEN (index-expression, selection-1 <,...selection-n>)

The CHOOSEN function is similar to the CHOOSEC function except that CHOOSEN returns a numeric value while
CHOOSEC returns a character value.

Index-expression value is to select from the arguments that follow, selection-1 to selection-n are the numeric values
from which to choose. If the first argument is negative, the function counts backwards from the list of arguments, and
returns that value.

EXAMPLE:

data _null_;
ITEMNUMBER = choosen(1,200,250,78,48,79);
RANK = choosen(-2,1,2,3,4,5);
SCORE = choosen(3,220,187,197,302,234);
VALUE = choosen(-5,-20,34,45,-90,45);
put ITEMNUMBER= RANK= SCORE= VALUE= ;
run;

SAS writes the following line to the log: ITEMNUMBER=200 RANK=4 SCORE=197 VALUE=-20.

CONCLUSION:

7
These new functions are valuable additions to existing SAS functions. Efficient and timely usage of these functions
makes the code elegant, simple and easy to maintain.

REFERENCES:

SAS Institute, Inc. SAS(R) 9.2 Language Reference: Dictionary, Fourth Edition. Online documentation:

IFC Function:
http://support.sas.com/documentation/cdl/en/lrdict/64316/HTML/default/viewer.htm#a002604573.htm

IFN Function:
http://support.sas.com/documentation/cdl/en/lrdict/64316/HTML/default/viewer.htm#a002604573.htm

CHOOSEC Function:
http://support.sas.com/documentation/cdl/en/lrdict/64316/HTML/default/viewer.htm#a002604617.htm

CHOOSEN Function:
http://support.sas.com/documentation/cdl/en/lrdict/64316/HTML/default/viewer.htm#a002604619.htm

ACKNOWLEDGMENTS

I would like to thank my management and friends at GCE Solutions, for motivating me to write this paper
and inspiring me to present at the conference.

CONTACT INFORMATION

Your comments and questions are valued and encouraged. Contact the author at:

Name: Geetha Kesireddi

Enterprise: GCE Solutions
Work Phone: 040-40273548
Email: geetha.kesireddy@gcesolutions.com
Website: https://gcesolutions.com
SAS® and all other SAS®

Institute Inc. product or service names are registered trademarks or trademarks of SAS®
Institute Inc. in the USA and other countries. ®indicates USA registration.

Other brand and product names are trademarks of their respective companies.