Data with semicolons

1
Chapter 13
2
Data with semicolons

• If we try to read this data, SAS will definitely
• complain.
• data new
• input author 15. title 30.
• datalines
• Smith, Alan Everything you want to know about
  
• Davis, Greg ?,!,etc.
• proc print
• run

3
datalines4

• data new
• input author 15. title 30.
• datalines4
• Smith, Alan Everything you want to know about
  
• Davis, Greg ?,!,etc.
• proc print
• run

4
Reading from a text file

• We saw in Chapter 12 how we could read data
• from a textfile.
• data new
• infile "Csomefile.txt"
• input x y z
• run
• proc print
• run

5

• We could also give the file a name first.
• data new
• filename whatever "Csomefile.txt"
• infile whatever
• input x y z
• run
• proc print
• run

6
Missover

• Data
• Male 2 8
• Female 3
• Female 4 1

• data new
• filename whatever "Csomefile.txt"
• infile whatever missover
• input x y z
• run
• proc print
• run

7
Reading Multiple Files

• Suppose we have 3 files examscores1.txt,
• examscores2.txt and examscores3.txt
• data new
• filename whatever "Cexamscores.txt"
• infile whatever
• input x y z
• run
• Each file contains 3 columns of numbers.

8

• data new
• filename whatever ("Cfile01.txt"
  "Cdata01.txt")
• infile whatever
• input x y z
• run
• Again, all files must have the same structure.

9
Writing ASCII data to an external file

• data new
• infile "Cinputfile.txt"
• file "Coutputfile.txt"
• input x y z
• total x y z
• meanxy(xy)/2
• put x y z total meanxy
• run

10
Writing CSV files

• options missing" "
• data new
• input x y z
• datalines
• 2 2 1
• 1 0 8
• ods listing close
• ods csv file"Cnewfile.csv"
• proc print datanew
• run
• ods csv close
• ods listing

11
Permanent SAS data set

• A permanent SAS data set can only be used by
• SAS. The authors state that there is a free
• SAS viewer program to view and print SAS
• data sets.
• These datasets will usually use more storage
• than the original raw data. The biggest reason
• to use SAS data sets is speed.

12

• If you plan to run many different analyses on
• a dataset that will not be changing, its a good
• idea to make a permanent data set.
• SAS data sets can easily be transferred to
• other SAS users. You dont need to know the
• structure of the data since the variables, labels
• and formats have already been defined.

13
Creating a SAS data set

• libname mydata "C\SASDATA"
• data mydata.file2
• input x y z
• datalines
• 2 2 1
• 1 0 8
• run

14
The file that was just created.
15
Reading SAS data sets

• libname abc "C\SASDATA"
• proc print dataabc.file2
• proc means dataabc.file2
• var x
• run
• In the procedures you must use the data
• option. Otherwise, SAS will not know what
• dataset you are referring to.

16
(No Transcript)
17
Proc Contents

• This procedure will list information about the
• SAS dataset especially the variables and
• their type.
• libname abc "C\SASDATA"
• proc contents dataabc.file2 varnum
• run

18
(No Transcript)
19
SAS data sets with formats

• If you have created a SAS data set that creates
• user-created formats to variables, you must
• make the format library permanent as well.
• If you send someone the SAS data set, be sure
• to send them the format library.

20
Creating the format library

• libname mydata "C\SASDATA"
• options fmtsearch (mydata)
• proc format librarymydata
• value ggroup mmale
• ffemale
• run
• data mydata.file2
• input gender y z
• format group ggroup
• datalines
• m 2 1
• f 0 8
• run

21
A file containing the formats.
22
Read SAS data sets with formats

• libname newdata "C\SASDATA"
• options fmtsearch(newdata)
• proc print datanewdata.file2
• run

23
Working with large data sets

• The authors have suggestions for working more
• efficiently with large data sets.
• On a PC, a dataset with 500,000
• observations
• and 50 observations might be considered large.
• On a mainframe, a dataset with millions of
• observations might be considered large.

24
Dont read files unnecessarily

• Inefficient
• libname newdata "C\SASDATA"
• data mine
• set newdata.file01
• run
• proc print datamine
• run
• Efficient
• libname newdata "C\SASDATA"
• proc print datanewdata.file01
• run

25
DROP unnecessary variables

• Inefficient
• data mine
• input week1-week52
• avgtempmean(of week1-week52)
• run
• Efficient
• data mine
• input week1-week52
• avgtempmean(of week1-week52)
• drop (week1-week52)
• run

26
Use DROP(or KEEP) with SET.

• Inefficient
• data mine
• set old
• drop x1-x50 y z1-z20
• etc
• Efficient
• data mine
• set old (dropx1-x50 y z1-z20)
• etc

27
Use CLASS instead of BY

• Inefficient
• proc sort datanew
• by gender
• proc means datanew
• by gender
• var age gpa
• Efficient
• proc means datanew
• class gender
• var age gpa

28
Use WHERE instead of IF

• Inefficient
• data new
• set old
• if x ge 20
• run

• Efficient
• data new
• set old
• where x ge 20
• run
• Or
• data new
• set old (where (x ge 20))
• run

29
Use WHERE in procedures

• Inefficient
• data new
• set old
• where x ge 20
• run
• proc means datanew
• var age gpa
• run

• Efficient
• proc means dataold
• where x ge 20
• var age gpa
• run
• Or
• proc means
• dataold(where (x ge 20))
• var age gpa
• run

30
Use ELSE IF instead of multiple IFs

• Inefficient
• data new
• input crsavg
• if crsavg ge 90 then LG"A"
• if 80 le crsavg lt 90 then LG"B"
• if 70 le crsavg lt 80 then LG"C"
• if 60 le crsavg lt 70 then LG"D"
• if crsavg lt 60 then LG"F"
• run

31

• Efficient
• data new
• input crsavg
• if crsavg ge 90 then LG"A"
• else if 80 le crsavg lt 90 then LG"B"
• else if 70 le crsavg lt 80 then LG"C"
• else if 60 le crsavg lt 70 then LG"D"
• else if crsavg lt 60 then LG"F"
• run

32

• When using multiple IFs, its best to make the
• first IF the one that is most likely to be true.

33
Save Summary Statistics

• This will be useful if you plan to do further
• analysis on these results.
• libname newdata "C\SASDATA"
• proc means datanewdata.file01
• class gender
• var age height weight
• output outnewdata.means meanage height weight
• run

34
Use only the first n obs to test code.

• data test
• set hugedata(obs10)
• etc.
• run
• OR
• proc print datahugedata(obs10)
• run

35
Adding few obs to a large dataset

• proc append basehugedata datanewfile
• run