Fifty Ways to Lose Your Data (and How to Avoid Them)
Fifty Ways to Lose Your Data (and How to Avoid Them)
Linda Jolley and Jane Stroupe, SAS Institute Inc., Cary, NC
This paper groups possible destructive techniques and their
solutions into the following categories:
NOVICE PROBLEMS
Many errors that beginning SAS® programmers encounter fall into other categories; however, three are identified as
common misconceptions.
1. Assuming data must be created before it can be used.
data perm.DataSet;
proc print data = perm.DataSet;
run;
Solution: Refer to the permanent data set in the PRINT procedure. It is not necessary to have a DATA step in
order to use a data set name in the PROC statement.
proc print data = perm.DataSet;
run;
2. Not keeping all of the variables that you want to keep.
data NewData;
set OldData; /*The data set OldData contains the variables X, Y, and Z. */
keep Z; /*The only variables kept should be the variables W and Z. */
W = X + Y;
run;
Solution: Specify all of the variables you want to keep (even new ones that are created in the DATA step) in the
KEEP statement.
data NewData;
set OldData;
keep Z W;
W = X + Y;
run;
SAS Global Forum 2009 Foundations and Fundamentals
2
3. Not knowing the names of your permanent data sets and creating a new one with the name of an existing
data set.
data perm.old_data;
set other_data;
if x = . then x = 5;
run;
Solution: Look at the names of the existing permanent data sets before creating a new one.
proc datasets lib = perm nods;
contents data = _all_;
run;
data perm.new_name;
set other_data;
if x = . then x = 5;
run;
NOT KNOWING YOUR DATA
The first rule of programming should be 'know thy data.' Unfortunately, many programmers rush into their
programming tasks without considering the values of their data. The following are examples of ignoring that rule.
4. Not knowing the case of character data.
data you;
set perm.Names;
where FirstName contains 'il';
run;
Solution: Use the UPCASE function when comparing character variables with constants.
data you;
set perm.Names;
where UPCASE(FirstName) contains 'IL';
run;
5. Incorrectly choosing the length of numeric data.
data five;
length num5 5 num8 8;
do num8 = 10000000000 to 10000000000000 by 100000000000;
num5 = num8;
output;
end;
run;
Here are the observations six through eight in the data set FIVE. Notice that the values of NUM5 and NUM8 are
not equal starting with observation seven, even though the assignment statement NUM5=NUM8 makes them
equal.
Obs num5 num8
6 510000000000 510000000000
7 609999998976 610000000000
8 709999998976 710000000000
SAS Global Forum 2009 Foundations and Fundamentals
3
Solution: Do not change the length of numeric data if the values are large numbers.
data five;
do num8 = 10000000000 to 10000000000000 by 100000000000;
num5 = num8;
output;
end;
run;
6. Incorrectly determining the length of character data.
trainer.dat
Linda Jolley, Technical Trainer
Jane Stroupe, Technical Trainer
data instructor;
infile 'trainer.dat' dlm = ',';
input Name $ Job;
run;
The resulting data follows:
Name Job
Linda Jo Technica
Jane Str Technica
Solution: When using list input, specify the length of the character data. List input is used to read this data since
it is free format data. The default length of character data is 8, which can sometimes result in data truncation.
data instructor;
length Name $ 12 Job $ 17;
infile 'trainer.dat' dlm = ',';
input A $ B $;
run;
Alternate Solution: Use the colon modifier to specify an informat in the INPUT statement.
data instructor;
infile datalines dlm = ',';
input Name : $12. Job : $17.;
run;
7. Having large numeric data that results in precision problems.
data BigNumber_Num;
input Num;
datalines;
12345678901234567890
;
run;
Numeric data can store 16 or 17 significant digits in a length of 8; therefore, if a numeric value has more than 16
or 17 digits, you lose precision. The resulting value of NUM in this example is 12345678901234567168.
Solution: Create a character value instead of a numeric value. The value of this character variable would be
12345678901234567890.
data BigNumber_Char;
input Num : $20.;
datalines;
12345678901234567890
;
run;
SAS Global Forum 2009 Foundations and Fundamentals
4
8. Creating a numeric variable from a character variable.
data Try_Again;
set BigNumber_Char;
NewNum = input(Num, 20.);
format NewNum 20.;
run;
Solution: The problem continues to be the length of the original number. Even if you create a character value
instead of a numeric one for the variable NUM and then attempt to use the INPUT function to create a numeric
variable from the character value, you lose precision.
9. Using PROC IMPORT to determine the type of variable.
proc import out = work.teens datafile = "teens.xls"
dbms = excel replace;
run;
The Excel spreadsheet:
Solution: Use the MIXED=YES option in the PROC IMPORT statement. The MIXED=YES option assigns a SAS
character type for the column and converts all numeric data values to character data values when mixed data
types are found.
proc import out = work.teens
datafile = "teens.xls"
dbms = excel replace;
mixed=yes;
run;
10. Having raw data with long record lengths.
data LongRecl;
infile 'LongRec.dat';
input Char $480.;
Len = length(Char);
run;
Solution: Use the LRECL= option in the INFILE statement since the record length is longer than the default (256
bytes in Windows and UNIX).
data LongRecl;
infile 'LongRec.dat' lrecl = 480;
input Char $480.;
Len = length(Char);
run;
11. Not knowing that the raw data has varying lengths.
numbers.dat
143
36921
2
4509
SAS Global Forum 2009 Foundations and Fundamentals
5
data varying;
infile 'numbers.dat';
input Num 6.;
run;
The result of the DATA step follows:
Obs Num
1 36921
2 4509
Solution: See Number 12.
12. Not knowing that the raw data has varying lengths (Version 2).
data varying;
infile 'numbers.dat' missover;
input Num 6.;
run;
The result of the DATA step follows:
Obs Num
1 .
2 .
3 .
4 .
Solution to both: Use the TRUNCOVER option in the INFILE statement.
data varying;
infile 'numbers.dat' truncover;
input Num 6.;
run;
The result of the DATA step follows:
Obs Num
1 143
2 36921
3 2
4 4509
13. Having multiple delimiters treated as one delimiter.
fullnames.dat
Thomas, John Marshall
Sherrill, Franklin Woodford
data multiple;
infile 'fullnames.dat' dlm = ', ';
input LastName : $8. FirstMName : $25.;
run;
SAS Global Forum 2009 Foundations and Fundamentals
6
The result of this DATA step follows:
First
Obs LastName MName
1 Thomas John
2 Sherrill Franklin
Solution: Use the DLMSTR= option instead of the DLM= option in the INFILE statement. This SAS 9.2 option
specifies that the delimiter must be the exact string. In this example, the delimiter must be a comma followed by
a blank.
data multiple;
infile 'fullnames.dat' dlmstr = ', ';
input LastName : $8. FirstMName : $25.;
run;
The result of this DATA step follows:
Obs LastName FirstMName
1 Thomas John Marshall
2 Sherrill Franklin Woodford
14. Having class variables that are missing when creating statistics on an analysis variable.
proc summary data = perm.class nway;
output out = summary mean = AvgAge;
class Group;
var Age;
run;
For most procedures that calculate statistics, missing CLASS variables are not included in the output data sets.
The data used for this example is
Obs Name Age Group
1 Michelle 15
2 Steve 15
3 James 12 A
4 Jeffrey 13 A
5 John 12 A
20 Louise 12 D
21 Mary 15 D
The resulting data created by the SUMMARY procedure is as follows:
Obs Group _TYPE_ _FREQ_ AvgAge
1 A 1 4 12.0000
2 B 1 6 14.3333
3 C 1 1 11.0000
4 D 1 8 13.5000
SAS Global Forum 2009 Foundations and Fundamentals
7
Solution: Use the MISSING option in the PROC SUMMARY statement.
proc summary data = perm.class missing nway;
output out = summary mean = AvgAge;
class Group;
var Age;
run;
The resulting data created by this SUMMARY procedure is as follows:
Obs Group _TYPE_ _FREQ_ AvgAge
1 1 2 15.0000
2 A 1 4 12.0000
3 B 1 6 14.3333
4 C 1 1 11.0000
5 D 1 8 13.5000
SYNTAX ERRORS
Often syntax errors stop execution of a step; therefore, a DATA step might not remove data from your data set.
However, there are some syntax errors that have a destructive behavior. Watch out for them!
15. Using the SQL procedure with too many semicolons.
proc sql;
delete from test;
where name='John';
quit;
Although this PROC SQL step causes a syntax error, it is too late. All of the rows of data have been deleted.
Solution: Put a semicolon only at the end of the SQL DELETE statement.
proc sql;
delete from test
where name='John';
quit;
Alternate Solution: Use the NOEXEC option in the PROC SQL statement. This option checks the syntax of a
PROC SQL step without actually executing it.
proc sql noexec;
delete from test
where name='John';
quit;
16. Omitting a semicolon and misspelling a keyword.
data two
ser perm.one;
run;
This program creates three data sets: TWO, SER, and PERM.ONE. Since there are no other statements in the
DATA step, the data set PERM.ONE is created with no observations.
Solution: Add a semicolon at the end of the DATA statement and/or spell SET correctly. (The default SAS
system option DATASTMTCHK = COREKEYWORDS prohibits the SET keyword as a one-level SAS data set
name.)
data two;
set perm.one;
run;
OR
SAS Global Forum 2009 Foundations and Fundamentals
8
data two
set perm.one;
run;
17. Hitting the incorrect key on the keyboard causing more misspellings.
proc sql;
delect * from perm.one;
quit;
Log
1171 proc sql;
1172 delect * from one;
------
1
WARNING 1-322: Assuming the symbol DELETE was misspelled as delect.
NOTE: 3 rows were deleted from WORK.ONE.
1173 quit;
Solution: Watch your fingers as you type. SQL interprets DELECT as DELETE and deletes all the rows of data
from the data set PERM.ONE. The 'S' and the 'D' are next to each other on a standard keyboard.
proc sql;
select * from one;
quit;
18. Merging without a BY statement.
data set CHAR1 data set CHAR2
one two one three
A B A X
C D B Y
E F C Z
data char3;
merge char1 char2;
run;
Without a BY statement, the merge combines the first observations of each data set, the second observations,
and so forth. Therefore, there is incorrect data, and much of it is lost. This is the resulting data.
Obs one two three
1 A B X
2 B D Y
3 C F Z
Solution: Use a BY statement.
data char3;
merge char1 char2;
by a; /* the input data is already sorted */
run;
SAS Global Forum 2009 Foundations and Fundamentals
9
The resulting data set now contains the correct data:
Obs one two three
1 A B X
2 B Y
3 C D Z
19. Merging on one common variable when there is more than one common variable.
data set CHAR1 data set CHAR4
one two one two
A B A X
C D B Y
E F C Z
data char5;
merge char1 char4;
by one; /* the data is already sorted */
run;
When the two data sets are merged only on the variable A, the values of B from the data set CHAR4 overwrite
the values of B from the data set CHAR1. The resulting data set follows:
Obs one two
1 A X
2 B Y
3 C Z
Solution: Merge on all like-named variables.
data char6;
merge char1 char4;
by one two; /* the data is already sorted */
run;
The resulting data follows:
Obs one two
1 A B
2 A X
3 B Y
4 C D
5 C Z
6 E F
Alternate Solution: Rename all of the like-named variables except the variable on which the merge should be
done.
data char7;
merge char1 char4(rename=(Two=NewTwo));
by one; /* the data is already sorted */
run;
SAS Global Forum 2009 Foundations and Fundamentals
10
The resulting data follows:
New
Obs one two Two
1 A B X
2 B Y
3 C D Z
4 E F
20. Forgetting to reset an option.
options obs = 10;
proc print data = perm.one;
run;
proc sort data = perm.one;
by x;
run;
Solution: Reset the option. In the case of the OBS = option, use OBS = MAX to read all of the data.
options obs = max;
proc sort data = perm.one;
by x;
run;
21. Reading in delimited data incorrectly.
locations.dat
Libertyville, Lake, IL
Kansas City,,KS
data locations;
infile 'locations.dat';
input City : $12. County : $10. State : $2.;
run;
The problem here is that no delimiter has been specified; therefore, the default delimiter of a blank is being used.
The resulting data set follows:
Obs City County State
1 Libertyville Lake, IL
Solution: See Number 23.
22. Reading in delimited data incorrectly (Version 2).
data locations;
infile 'locations.dat' dlm = ',';
input City : $12. County : $10. State : $2.;
run;
This is a good start to fixing the problem. However, multiple delimiters are treated as one delimiter, causing the
INPUT statement to reach the end of the line in the second observation. The resulting data set is no different
than it was for the previous example.
Solution: See Number 23.
SAS Global Forum 2009 Foundations and Fundamentals
11
23. Reading in delimited data incorrectly (Version 3).
data locations;
infile 'locations.dat' dlm = ',' missover;
input City : $12. County : $10. State : $2.;
run;
The MISSOVER option prevents the INPUT statement from reaching the end of the line. However, for the
second observation, the value of COUNTY is read in as KS. The resulting data follows:
Obs City County State
1 Libertyville Lake IL
2 Kansas City KS
Solution to all three: Use the DSD option in the INFILE statement.
data locations;
infile 'locations.dat' dlm = ',' missover dsd;
input City : $12. County : $10. State : $2.;
run;
The DSD option (Delimiter-Sensitive Data) uses a comma as the default delimiter, treats consecutive commas as
a missing value, and reads a character value enclosed in quotation marks and removes the quotation marks from
the character string before the value is stored. The resulting data set follows:
Obs City County State
1 Libertyville Lake IL
2 Kansas City KS
24. Reading mixed record types incorrectly.
holiday.dat
US,Independence Day,04Jul2009
FR,Bastille Day,14-7-2009
data Holiday;
infile 'holiday.dat' dsd;
input Country : $2. Holiday : $16.;
if Country = 'US' then input Date : date9.;
else input Date : ddmmyy10.;
format date mmddyy10.;
run;
The resulting data follows:
Obs Country Holiday Date
1 US Independence Day .
Solution: Use multiple INPUT statements with a single trailing @ sign and check a condition before reading
more data. The single trailing @ sign holds the current record in the input buffer so that the second INPUT
statement can continue to read from that record.
data Holiday;
infile 'holiday.dat' dsd;
input Country : $2. Holiday : $16. @;
if Country = 'US' then input Date : date9.;
else input Date : ddmmyy10.;
format date mmddyy10.;
run;
SAS Global Forum 2009 Foundations and Fundamentals
12
Alternate Solution: Use the DATASTYLE= SAS system option and the ANYDTDTE informat on the Date
variable. The DATESTYLE= SAS system option identifies the order of month, day, and year. The default value of
LOCALE uses the style that has been specified by the LOCALE= system option. The ANYDTDTE informat reads
dates in various informats.
options datestyle = locale;
data Holiday;
infile 'holiday.dat' dsd;
input Country : $2. Holiday : $16. Date : anydtdte10.;
format date mmddyy10.;
run;
25. Using the NODUPKEY option with PROC SORT and losing the duplicate observations.
data test;
infile datalines @@;
input Name $;
datalines;
John Bill Sam Bill Tom Joe Chris Bill John Kent Chris
;
run;
proc sort data = test nodupkey;
by Name;
run;
Solution: Use the DUPOUT= option in the PROC SORT statement and name a data set that contains the
duplicate observations.
proc sort data = test nodupkey dupout = extras;
run;
26. Merging to make changes to data.
data set Grocery data set Update
ID Name Job Salary ID Name Job Salary
1065 Bob Clerk 35090 1065 Manager 45125
1112 Sue Manager 55890 1117 33589
1117 Sam Bagger 31298 1350 Bagger
1350 Peg Stocker 25897 1350 30876
data updated_grocery;
merge grocery update;
by idnum;
run;
Because the MERGE processes data sequentially and the like-named variables in the data set Update overwrite
the variables from Grocery, the resulting data set is improperly updated. The resulting data set follows.
Obs ID Name Job Salary
1 1065 Manager 45125
2 1112 Sue Manager 55890
3 1117 33589
4 1350 Bagger
5 1350 30876
SAS Global Forum 2009 Foundations and Fundamentals
13
Solution: Use the UPDATE statement instead of the MERGE statement.
data updated_grocery;
update grocery update;
by idnum;
run;
The UPDATE statement changes the values of selected observations in a master file by applying transactions. If
the data in the transaction data set is missing, the UPDATE statement does not overwrite existing values in the
master data set with missing ones in the transaction data set. The resulting data set follows.
Obs ID Name Job Salary
1 1065 Bob Manager 45125
2 1112 Sue Manager 55890
3 1117 Sam Bagger 33589
4 1350 Peg Bagger 30876
27. Reading hierarchical files and losing the last observation.
pets.dat
1 Michelle
Max cat
Fuschia cat
2 Theresa
Paisley dog
Sitka dog
3 Jane
4 Nancy
Winston cat
data pets;
retain Name;
keep Name NumDogs NumCats NumOther;
infile 'pets.dat';
input @1 FirstChar : $1. @;
if FirstChar in ('1','2','3','4','5','6','7','8','9') then do;
if _n_ ne 1 then output;
input Name : $8.;
NumDogs = 0;
NumCats = 0;
NumOther = 0;
end;
else do;
input @1 PetName : $15. Pet : $3.;
if upcase(Pet) = 'DOG' then NumDogs + 1;
else if upcase(Pet) = 'CAT' then NumCats + 1;
else NumOther + 1;
end;
run;
The resulting data set follows.
Num Num Num
Obs Name Dogs Cats Other
1 Michelle 0 2 0
2 Theresa 2 0 0
3 Jane 0 0 0
SAS Global Forum 2009 Foundations and Fundamentals
14
Solution: Specify the END= option in the INFILE statement and use conditional logic to output the last
observation.
data pets;
retain Name;
keep Name NumDogs NumCats NumOther;
infile 'pets.dat' end = LastObs;
input @1 FirstChar : $1. @;
if FirstChar in ('1','2','3','4','5','6','7','8','9') then do;
if _n_ ne 1 then output;
input Name : $8.;
NumDogs = 0;
NumCats = 0;
NumOther = 0;
end;
else do;
input @1 PetName : $15. Pet : $3.;
if upcase(Pet) = 'DOG' then NumDogs + 1;
else if upcase(Pet) = 'CAT' then NumCats + 1;
else NumOther + 1;
end;
if LastObs then output;
run;
The resulting data set follows.
Num Num Num
Obs Name Dogs Cats Other
1 Michelle 0 2 0
2 Theresa 2 0 0
3 Jane 0 0 0
4 Nancy 0 1 0
28. Allowing automatic conversion of numeric to character values.
Data set Numbers
NumValue
1234
5498
5698
data character;
set Numbers;
CharValue = substr(NumValue, 1, 4);
run;
When automatic conversion of numeric data to character values occurs, the digits are written out in 12 spaces
and are right aligned; therefore, the first four characters would be blank. The resulting data set follows.
Obs NumValue CharValue
1 1234
2 5498
3 5698
SAS Global Forum 2009 Foundations and Fundamentals
15
Alternate Solution: Use the PUT function with a specific number of digits if you know the number of digits.
data character;
set Numbers;
CharValue = substr(put(NumValue, 4.), 1, 4);
run;
Alternate Solution: Use LEFT function along with the PUT function with a specific number of digits if you do not
know the number of digits
data character;
set Numbers;
CharValue = substr(left(put(NumValue, 4.)), 1, 4);
run;
29. Not specifying the KEY variable as a DATA item when using the hash object with the OUTPUT method.
data _null_ ;
length List $100;
if _n_ = 1 then do;
declare hash ph();
ph.DefineKey('Date');
ph.DefineData('List');
ph.DefineDone();
end;
set PhoneCalls end = last;
if ph.find() ne 0 then do;
List = PhoneNum;
ph.add();
end;
else do;
List = catx(', ',List, PhoneNum);
ph.replace();
end;
if last then ph.output(dataset:'PhoneList');
run;
A partial listing of the data set PhoneCalls follows:
Obs Date PhoneNum
1 21DEC2006 324-6674
2 21DEC2006 312-5098
3 21DEC2006 312-9088
4 21DEC2006 324-3452
5 21DEC2006 312-7483
6 21DEC2006 324-8943
7 21DEC2006 312-7456
The variable DATE is not in the resulting data set.
Obs List
1 324-6674, 312-5098, 312-9088, 324-3452, 312-7483, 324-8943, 312-7456, 324-4544
2 312-5098, 312-4755, 389-0098, 324-6674
3 312-9088, 324-3452, 312-7456, 312-5098, 324-8943, 312-5098
SAS Global Forum 2009 Foundations and Fundamentals
16
Solution: Add the KEY variable as one of the DATA columns in the DEFINEDATA method. By default, the KEY
variable is not included as part of the data.
data _null_ ;
length List $100;
if _n_ = 1 then do;
declare hash ph();
ph.DefineKey('Date');
ph.DefineData('Date', 'List');
ph.DefineDone();
end;
set PhoneCalls end = last;
if ph.find() ne 0 then do;
List = PhoneNum;
ph.add();
end;
else do;
List = catx(', ',List, PhoneNum);
ph.replace();
end;
if last then ph.output(dataset:'PhoneList');
run;
30. Having duplicate keys when using a hash object. Only the first one is loaded into the hash object.
data all_inquiries;
drop rc;
length Product_ID 8;
if _N_ = 1 then do;
declare hash h(dataset:'orders');
h.defineKey('Order_ID');
h.defineData('Order_ID', 'Product_ID');
h.defineDone();
call missing(Product_ID);
end;
set inquiries;
rc=h.find();
run;
Solution: Use the MULTIDATA : ‘YES’ parameter in the DECLARE statement. MULTIDATA is a SAS 9.2
enhancement to the hash object syntax which allows multiple data items for each non-unique key.
data all_inquiries;
drop rc;
length Product_ID 8;
if _N_ = 1 then do;
declare hash h(dataset:'orders', multidata:'yes');
h.defineKey('Order_ID');
h.defineData('Order_ID', 'Product_ID');
h.defineDone();
call missing(Product_ID);
end;
set inquiries;
rc = h.find();
do while (rc = 0);
output;
rc=h.find_next();
end;
run;
SAS Global Forum 2009 Foundations and Fundamentals
17
31. Not including automatic variables in your data.
proc sort data = perm.class out = class;
by Group;
run;
data Summary;
set class nobs = NumObs;
keep Group AvgAge TotAge NumObs Count;
by Group;
if first.Group then do;
TotAge = 0;
Count = 0;
end;
TotAge + Age;
Count + 1;
if last.Group then do;
AvgAge = TotAge / Count;
output;
end;
run;
Solution: Create a DATA step variable that contains the value of the NOBS= variable, NUMOBS.
data Summary;
set class nobs = NumObs;
keep Group AvgAge TotAge TotObs Count;
TotObs = NumObs;
by Group;
if first.Group then do;
TotAge = 0;
Count = 0;
end;
TotAge + Age;
Count + 1;
if last.Group then do;
AvgAge = TotAge / Count;
output;
end;
run;
32. Reading from multiple raw data files in one DATA step.
There are three raw data files with the same fields but a different number of records. The DATA step stops when
execution reaches the end of file for the raw data file with the least number of records. Therefore, the DATA step
never reads the rest of the other raw data files.
Bill 32
John 38
Beth 33
Sue 35
data names;
infile 'names1.dat';
input Name $ Age;
output;
infile 'names2.dat';
input Name $ Age;
output;
infile 'names3.dat';
input Name $ Age;
output;
run;
SAS Global Forum 2009 Foundations and Fundamentals
18
Solution: Use the FILEVAR = option in the INFILE statement to specify a variable whose change in value
causes the INFILE statement to close the current input file and open a new one. In addition, use a DO WHILE
loop to read all of the records of each raw data file.
data routes;
length Readit $10;
input ReadIt;
infile in filevar = ReadIt end = LastFile pad;
do while(LastFile = 0);
input Name $ Age;
output;
end;
datalines;
names1.dat
names2.dat
names3.dat
;
run;
BAD LOGIC
33. Having a never met condition on a WHERE statement.
proc sort data = one;
by x;
where x = 12;
run;
Solution: Create a data set using the OUT= option. In this case, the data set TWO would be empty, but the
integrity of the data set ONE is maintained.
proc sort data = one out = two;
by x;
where x = 12;
run;
34. Omitting a condition on an IF statement.
data one;
set one;
if x then delete;
run;
Solution: By not specifying a value for the variable X, the condition IF X is true as long as X is not 0 and X is not
missing. Therefore, all of the observations with a nonzero value of X are deleted.
data one;
set one;
if x = 12 then delete;
run;
35. Specifying a false condition when using the DATA step with a MODIFY statement and a WHERE
statement.
data one;
modify one;
x = x * 2;
where x = 12;
run;
SAS Global Forum 2009 Foundations and Fundamentals
19
Solution: Use the IF statement instead of the WHERE statement.
data one;
modify one;
x = x * 2;
if x = 12;
run;
36. Omitting a LENGTH statement.
data char;
input a $ b $;
datalines;
ONE TWO
THREE FOUR
FIVE SIX
SEVEN EIGHT
NINE TEN
;
data NewChar;
set char;
A = A||B;
run;
Solution: Use a LENGTH statement before the SET statement.
data NewChar;
length A $ 16;
set char;
A = A||B;
run;
37. Using the SCAN function with multiple delimiters.
data test;
input pets : $20.;
pet2 = scan(pets, 2, ',');
datalines;
cat,dog,fish
dog,,bird
;
run;
Solution: By default, the SCAN function treats multiple delimiters as one delimiter. In addition, any delimiters
before or after the string are ignored. Use the 'M' option as the fourth option in the SCAN function (a 9.2
enhancement). It specifies that multiple consecutive delimiters (and any at the beginning or end of the first
argument) are treated as missing values between the delimiters; therefore, the variable being created has a
missing value.
data test;
input pets : $20.;
pet2=scan(pets, 2, ',', 'M');
datalines;
cat,dog,fish
dog,,bird
;
run;
SAS Global Forum 2009 Foundations and Fundamentals
20
38. Creating a random sample with the RANUNI function when you need to recreate the same sample later.
data subset;
do I = 1 to 10;
x = int(num * ranuni(0));
set Seuss point = x nobs = num;
output;
end;
stop;
run;
Solution: Use a positive integer as the seed in the RANUNI function.
data subset;
do I = 1 to 10;
x = int(num * ranuni(12345));
set Seuss point = x nobs = num;
output;
end;
stop;
run;
INTENTIONALLY DELETING DATA SETS
39. Using the DATASETS procedure to delete the entire data set.
proc datasets lib = work;
delete one;
run;
40. Using the SQL procedure to drop a data set.
proc sql;
drop table one;
quit;
41. Using the Explorer window to delete a data set.
Right-click on a SAS data set in the Explorer window and select DELETE. Be careful because this does not put
the data set into the Windows recycle bin.
42. Using the VIEWTABLE window to delete rows from a SAS data set.
Double-click on the data set in the Explorer window.
From the Menu bar, select EDIT