proc print

proc print data=your.data (obs=n) noobs;
var name dx1 cost;
where dx1=“222”;
sum cost;
run;

Options
- VAR select variables
- NOOBS suppress observation numbers
- SUM column totals
  - numeric variables
- ID identify other than obs num
Conditioning / Logical Delimiters
- WHERE
- EQ, NE, GT, LT, GE, LE, IN

helpful hint

It’s usually a good idea to print out the first couple of lines of a data set after you create it

Use OBS= to limit the number of observations
- (obs=n) must be in parentheses, and must precede the noobs option

proc sort

PROC SORT data=your.dat OUT= sortdat;  
/*note the output data set  is in the   work library */  
BY  var1; /* will be sorted by var1 ascending */  
RUN;

sequence and group observations
- doesn’t produce output
  - check your log!
options
- BY variable by which you want to sort
- OUT create another, sorted, output data file
- DESCENDING default is ascending

simple subset analysis using by

PROC PRINT data=sortdat; /* note using the output sorted data set */  
BY var1;  
SUM num_var;  
RUN;

use PROC SORT then PROC PRINT
- e.g. subtotal your PROC PRINT output

subsetting using where

can be used with most any sas proc
where condition;
- where name = ‘smith’;
special operators
- contains returns exact matches
  - where name contains ‘ith’;
- is / is not missing
  - where name is missing;
- = *’string’ return text that sounds like some character string
  - where name=*’smith’;

titlen and footnoten

title ‘your title here’;

good habit to give each of your analyses a title.
Can have up to 10 lines in a title, e.g.
- title ‘First Line of Your Title’;
- title2 ‘Second Line of Your Title’;
- title3 ‘Third Line of Your Title’;
same titles and footnotes until you change them
- hierarchical - new number n cancels all subsequent ones

label

LABEL;  
var1=”name for variable 1” ;  
var2=”name for variable 2” ;  
var3=”name for variable 3” ;

informative or descriptive name for a variable
- otherwise SAS uses the name from data input statement
PROC step only for that session;
DATA step permanently attached.

format

proc print data=your.data;  
var charge;   
format charge dollar11.2;   
run;

print a variable using an existing format
the variable called ‘charge’ is printed as an 11 number dollar figure with 2 decimal points
doesn’t change source data, just its appearance on the page
- (here) just for thisPROC
general form ‘FORMAT format_name w.d
- w is width
- . is a required delimiter
- number after . is number of decimal points
  - e.g. date9. gives you 02Jan1960

proc format

proc format;  
value format\_name   
range 1=’label’  
range 2 = ‘label’;  
run;

Create your own format
To be distinguished from (simple) FORMAT
2-step process
- name and create your own format with PROC FORMAT
- apply format you created with FORMAT
format name
- no longer than 32 characters
  - ($ indicates character)
- can’t end in a number
  - (otherwise SAS thinks you’re delimiting a width)

combining proc format and format

proc format; /* first create the format with PROC FORMAT*/  
value gndrfmt 1=’Female’ /*note, no dot when *creating* the format*/  
              2=’Male’  
              0=’Unknown’;  
run;  

proc print data=your.data;     
    format gender gndrfmt. ; /\* apply the format, note dot\*/    
run;

gender is a numeric variable coded 1 for females and 2 for males
want to assign the word “Female” and “Male”

format libraries

PROC - format applies only to that session
DATA - format remains attached to that variable
- have to invoke or restate the proc format statement every time you want to run a procedure on that data.
Painful Lesson
- publicly available data sets, e.g. BRFSS, come with format libraries that can be difficult to use and apply
- nofmterroption after defining the data set allows you to use dataset without format library

format options

proc format;  
value boardfmt 
               low-49=’below’  
               50-99=’average’  
               100-high=’above average’;

(character) indicator variable based on (numeric) ranges

proc format;  
value \$grade  
               ‘A’=’good’;  
               ‘B’=’fair’;  
               ‘C’=’poor’;  
               ‘U’=’see instructor’;

non-numeric (character) variables

proc print data=ia.empdata;  
format final\_exam \$grade.;  
run;

*again note need for dot when applying the format*

using the data step to permanently assign variable attributes

data your.sasfile; /* data step creates new data file */     
infile ‘raw-data-file’; /*location of external file on your machine */  
input            
@1 var1 \$3.  /* input statement to read in variables */    
@4 date mmddyy8. /* informat for how to read in this variable */  
@12 var2 \$3.      
@15 var3 3.    
@18 var4 3.;    
format date mmddyy10.;/*format for how to display this variable */  
label var1=’ Variable One’ /* labels for how to name these variables */  
var2=‘Variable Two’  
var3=’Variable Three’;  
run;

ods (output delivery system)

ods html file="C:\myfolder\myfile.html" style=sasweb;  
proc print data=mylib.mydat;  
var age weight gender;  
where gender="F";  
sum weight;  
run;  
ods html close;

creates a variety of outputs, e.g. pdf
ODS HTML opens, closes and manages HTML files
- note need to explicitly specify the extension
- folder must already exist
- style= to invoke optional outputs
  - tip sasweb is a nice style for html
note have to ‘close’ ODS

review

LIBNAME – used to create a “shortcut” to an existing folder
INFORMAT – to assign an existing SAS format to a variable when inputting data
LABEL - informative name for variables
- up to a 256 character label for each variable
FORMAT – how levels of a variable are displayed
- PROC step – just for that procedure
- DATA step – permanent
PROC FORMAT – to create your own formats

MANIPULATING DATA SETS

use a data step to manipulate variables

DATA    myLib.myNewDataSet;  
SET    myLib.myOlsDataSet;  
/* additional statements manipulating the existing data set*/ 
RUN;

not only to read in and create a data set
also anytime you manipulate a data set to create new variables, subset data, merge, etc…
use a set statement
- as opposed to infile or input

creating or redefining variables

basic format: *new_variable = expression;
- new_var = 10; (makes it a constant)
- new_var = ‘ten’; (makes it a character constant)
- new_var = old_var * 10; etc…
mathematical operations are prioritized
- exponentiation and neg prefixes > multiplication division > addition and subtraction
- use parentheses to keep things straight
common operands: +, -, *, /, ** (exponentiation)
- can also use EQ, LE, GE, etc…

sas functions

automated procedures
- canned routines ready for use
basic form: functionName(argument, argument …)
- new_var = INT(old_var)
  - returns the integer portion of the value in the parenthesis
- day_var = DAY(date_var)*
  - returns the day names of dates
can put another function as the argument of a function
NB: functions ignore missing values
- will not compute a result
- may get a warning or error depending on function
(not easy) to find SAS functions associated with a procedure
- help SAS help and documentation - SAS products - Base SAS - SAS Language dictionary - dictionary of language elements - SAS data set options

combining sas data sets: concatenating

DATA data.new;   
SET data.oldset1 data.oldset2 data.oldset3;   
RUN;

simply adds or appends one data set to the next
caution missing or differing variables (columns)

combining sas data sets: merging

DATA  data.new;   
MERGE data.oldset1 data.oldset2;   
BY var1;

Substitute the word MERGE for SET and include a BY statement
will perform a one-to-one match of two data sets

about merging data sets

all files must be SORTED by the BY variable
- (not so for PROC SQL)
use same syntax for more than 2 data sets
- for a one to many merge, use BY variable unique to one but repeated in others
caution if same named variables in data sets, will be overwritten

the pain of merging

merging is a very powerful procedure
be certain your new data set contains the variables you want
- check your log
- check your data

merging trick to identify / exclude non-matching observations

data new;  
merge old1 (in=selected) old2;  
by var;  
if selected=1;  
run;

*in = * option that identifies the data set the observation comes from
creates temporary variable called indata1
indicates whether the variable came from the first-named data set
use this info with a subsetting IF statement
keep the observation (row) only if merges with observation in first-named data set

the pain of merging II: variables must match exactly

convert a character variable to a numeric variable : input()

data new_data; /* create new data set */
set old_data;
num_var=input(char_var,informat.); /* e.g. numeric informat 3. */
run;

convert a numeric variable to a character variable: put()

data new_data;  
set old_data;  
char_var = put(num\_var, informat.) /* e.g. character informat $3. */

rename a variable on the fly

data myNewData  
set myOldData1 myOldData2 (rename=(oldVar=newVar) );  
run;

USING CONDITIONAL EXPRESSIONS

if-then-else

IF icd9 = ‘666’ THEN dx= ‘damien’;  
ELSE dx = "john";

can use comparison operators: EQ, NE, GT, LT, GE
IF –THEN statements are case sensitive
include a concluding ELSE statement
- more efficient computationally and insures exclusive categories

using if-then-else to group variables

DATA new;  
SET old;  
IF cost = . THEN cat = ‘missing’;  
ELSE IF cost LT 2000 THEN cat = ‘low’;  
ELSE IF cost GT 2000 & cost LT 10,000 THEN cat = ‘mod’;  
ELSE cat = ‘high’;  
RUN;

creates a new categorical variable
compares the value of a variable to a list of variables

using an in statement to create a look up list

DATA new;  
SET old;  
IF dx IN (‘myocardial infarction’, ‘angina’, ‘chest pain’)  THEN cardiac= 1;  
ELSE cardiac=0;  
RUN;

if-then-do-end to group statements

IF condition THEN DO;  
action;  
action;  
action;  
END;

data new_data;  
set old_data;  
if var3=’X’ then total=sum(num_var1, num_var2);   
ratio =  num_var1 / num_var2;    
pct =  (num_var1 / total)*100  
run;

only the new variable “total” is tied to the conditional statement about var3
- “ratio” and “pct” will be calculated regardless of value for var3

data new_data;  
set old_data;  
total=sum(num_var1, num_var2);    
if var3=’X’ then do;  
ratio =  num_var1 / num_var2;  
pct =  (num_var1 / total)*100;    
end;  
else if var3=’Y’ then do;  
ratio =  num_var1 / num_var2;  
pct =  (num_var1 / total)*100;  
end;  
run;

execute multiple conditional statements based on a prior condition with a DO statement
stop the loop with an END statement
a way of grouping statements

arrays to create lists of variables

data new;  
set old;  
if dx1=. then dx1=999;  
if dx2=. then dx2=999;  
if dx3=. then dx3=999;  
.  
.  
.  
if dx100=. then dx100=999;   
if sex = . then sex = 999;   
if age = . then age = 999;   
run;

gets the job done, but slow and tedious

data new;  
set old;  
array alldxs[102] dx1-dx100 age sex;  /*create the array*/  
do i = 1 to 105;  /*create subscript (new variable) to index the array*/  
if alldxs[i]=. then alldxs[i]= 999;  /*apply statement to array*/  
end;   
drop i;/*drop the index variable from the data set*/  
run;

assign the variables to an array
apply the repetitive statement to the indexed array name
place in do loop so will repeat

using an array to restructure a data set

you have

|sub | obs1 | obs2 | obs3 |
|1 | 4 | 5 | 6 |
|2 | 7 | 8 | 9 |

you want

| sub | time | obs |
| 1 | 1 | 4 |
| 1 | 2 | 5 |
| 1 | 3 | 6 |
| 2 | 1 | 7 |
| 2 | 2 | 8 |
| 2 | 3 | 9 |

accomplish with an array

data new;
set old;
array myvars[3] obs1-obs3; /*define an array (list) of 3 existing variables*/  
do time=1 to 3;/*using time as the indexing variable*/  
obs=myvars[time];/*create variable "obs" equal to the indexed array variable*/    
output;/*add line of output to data set*/   
end;  
drop obs1-obs3;  
run;

note: “output” is implicit at end of every row (observation) in a data statement
here explicitly creating new observation that contains subject, time and obs

restricting observations with if and where statements

data females;  
set male_female  
if sex = ‘F’;  
run;

if, alone, restricts data to a subset by keeping observations that meet the criteria
- note: could also use a delete statement to do the same thing
  - e.g. IF sex = ‘M’ THEN DELETE;
WHERE vs IF
- WHERE best with PROC
- IF best with DATA

restricting variables with drop and keep statements

data new (drop=var1 var2);  
SET old;  
total=var3 + var4;  
run;

DROP=( ) , KEEP=( ) in a DATA step

UTILITY PROCEDURES