understanding sas data step processing
DESCRIPTION
Understanding SAS Data Step Processing. Alan C. Elliott stattutorials.com. Reading Raw Data. Using the following SAS program: DATA NEW; INPUT ID $ AGE TEMPC; TEMPF=TEMPC*( 9 / 5 )+ 32 ; DATALINES; 0001 24 37.3 0002 35 38.2 ; run ; proc print ; run ;. Overview of SAS Data Step. - PowerPoint PPT PresentationTRANSCRIPT
Understanding SAS Data Step Processing
Alan C. Elliottstattutorials.com
Alan C. Elliott, stattutorials.com
Reading Raw Data
• Using the following SAS program:DATA NEW;INPUT ID $ AGE TEMPC;TEMPF=TEMPC*(9/5)+32;DATALINES;0001 24 37.30002 35 38.2;run;proc print;run;
Alan C. Elliott, stattutorials.com
Overview of SAS Data Step
Compile Phase(Look at Syntax)
Execution Phase(Read data, Calculate)
Output Phase(Create Data Set)
Alan C. Elliott, stattutorials.com
Compile PhaseDATA NEW;INPUT ID $ AGE TEMPC;TEMPF=TEMPC*(9/5)+32;DATALINES;0001 24 37.30002 35 38.2;run;proc print;run;
SAS Checks the syntax of the program.• Identifies type and
length of each variable• Does any variable need
conversion?
If everything is okay, proceed to the next step.
If errors are discovered, SAS attempts to interpret what you mean. If SAS can’t correct the error, it prints an error message to the log.
Alan C. Elliott, stattutorials.com
Create Input Buffer• SAS creates an input buffer• INPUT BUFFER contains data as it is read in
DATALINES;0001 24 37.30002 35 38.2;
1 2 3 4 5 6 7 8 9 10 11 120 0 0 1 2 4 3 7 . 3
INPUT BUFFER
Alan C. Elliott, stattutorials.com
Execution Phase
• PROGRAM DATA VECTOR (PDV) is created and contains information about the variables
• Two automatic variables _N_ and _ERROR_ and a position for each of the four variables in the DATA step.
• Sets _N_ = 1 _ERROR_ = 0 (no initial error) and remaining variables to missing.
_N_ _ERROR_ ID AGE TEMPC TEMPF
1 0 . . .
Alan C. Elliott, stattutorials.com
Buffer to PDV1 2 3 4 5 6 7 8 9 10 11 12
0 0 0 1 2 4 3 7 . 3
_N_ _ERROR_ ID AGE TEMPC TEMPF
1 0 0001 24 37.3 .
Calculated value
Buffer
PDV
_N_ _ERROR_ ID AGE TEMPC TEMPF
1 0 0001 24 37.3 99.14
Processes the code TEMPF=TEMPC*(9/5)+32; Initially missing
Reads 1st record
If there is an executable statement…
Alan C. Elliott, stattutorials.com
Output Phase• The values in the PDV are written to the
output data set (NEW) as the first observation:
_N_ _ERROR_ ID AGE TEMPC TEMPF
1 0 0001 24 37.3 99.14
ID AGE TEMPC TEMPF
0001 24 37.3 99.14
This is the first record in the output data set
named “NEW.” Note that _N_ and
_ERROR_ are dropped.
From PDV
Write data to data set.
Alan C. Elliott, stattutorials.com
Exceptions to Missing in PDV
• Some data values are not initially set to missing in the PDV – variables in a RETAIN statement– variables created in a SUM statement– data elements in a _TEMPORARY_ array– variables created with options in the FILE or INFILE
statements• These exceptions are covered later.
_N_ _ERROR_ ID AGE TEMPC TEMPF
1 0 . . .
Initial values usually set to missing in PDV
Alan C. Elliott, stattutorials.com
Next data record read
• Once SAS finished reading the first data record, it continues the same process, and reads the second record…sending results to output data set (named NEW in this case.)
• …and so on for all records.
ID AGE TEMPC TEMPF
0001 24 37.3 99.14
0002 35 38.2 100.76
Alan C. Elliott, stattutorials.com
Descriptor Information
• For the data set, SAS creates and maintains a description about each SAS data set:– data set attributes– variable attributes– the name of the data set– member type, the date and time that the data set
was created, and the number, names and data types (character or numeric) of the variables.
Alan C. Elliott, stattutorials.com
Data Set Description
proc datasets ; contents data=new;run;
Contents output… (abbreviated)
# Name Member Type
File Size Last Modified
1 NEW DATA 5120 20Nov13:08:59:32
Alternate program
proc contents data= new; run;
Alan C. Elliott, stattutorials.com
Description output continued…Data Set Name WORK.NEW Observations 2Member Type DATA Variables 4Engine V9 Indexes 0Created Wed, Nov 20, 2013
08:59:32 AMObservation Length 32
Last Modified Wed, Nov 20, 2013 08:59:32 AM
Deleted Observations
0
Protection Compressed NOData Set Type Sorted NOLabelData Representation WINDOWS_64Encoding wlatin1 Western
(Windows)
Alan C. Elliott, stattutorials.com
Description output continued…
Alphabetic List of Variables and Attributes# Variable Type Len2 AGE Num 81 ID Char 83 TEMPC Num 84 TEMPF Num 8
Alan C. Elliott, stattutorials.com
Original ProgramDATA NEW;INPUT ID $ AGE TEMPC;TEMPF=TEMPC*(9/5)+32;DATALINES;0001 24 37.30002 35 38.2;run;proc print;run;
Alan C. Elliott, stattutorials.com
Original ProgramDATA NEW;INPUT ID $ AGE TEMPC;TEMPF=TEMPC*(9/5)+32;DATALINES;0001 24 37.30002 35 38.2;run;proc print;run;
Obs ID AGE TEMPC
TEMPF
1 0001 24 37.3 99.142 0002 35 38.2 100.76
Program output
Alan C. Elliott, stattutorials.com
Example of ErrorDATA NEW;INPUT ID $ AGE TEMPC;TEMPF=TEMPC*(9/5)+32DATALINES;0001 24 37.30002 35 38.2;run;proc print;run;
proc datasets ; contents data=new;run;
Missing Semi-colon
Alan C. Elliott, stattutorials.com
76 DATA NEW;77 INPUT ID $ AGE TEMPC;78 TEMPF=TEMPC*(9/5)+3279 DATALINES; --------- 2280 0001 24 37.3 ---- 180ERROR 22-322: Syntax error, expecting one of the following: !, !!, &, *, **, +, -, /, <, <=, <>, =, >, ><, >=, AND, EQ, GE, GT, IN, LE, LT, MAX, MIN, NE, NG, NL, NOTIN, OR, ^=, |, ||, ~=.
ERROR 180-322: Statement is not valid or it is used out of proper order.
81 0002 35 38.282 ;83 run;
ERROR: No DATALINES or INFILE statement.
Error found during compilation
Alan C. Elliott, stattutorials.com
Summary - Compilation Phase• During Compilation– Check syntax– Identify type and length of each new variable (is a data type
conversion needed?)– creates input buffer if there is an INPUT statement for an
external file – creates the Program Data Vector (PDV)– creates descriptor information for data sets and variable
attributes – Other options not discussed here: DROP; KEEP; RENAME;
RETAIN; WHERE; LABEL; LENGTH; FORMAT; ARRAY; BY; ATTRIB; END=, IN=, FIRST, LAST, POINT=
Alan C. Elliott, stattutorials.com
Summary – Execution Phase1. The DATA step iterates once for each observation being
created. 2. Each time the DATA statement executes, _N_ is incremented by
1.3. Newly created variables set to missing in the PDV.4. SAS reads a data record from a raw data file into the input
buffer (there are other possibilities not discussed here).5. SAS executes any other programming statements for the
current record.6. At the end of the data statements (RUN;) SAS writes an
observation to the SAS data set (OUTPUT PHASE)7. SAS returns to the top of the DATA step (Step 3 above)8. The DATA step terminates when there is no more data.
Alan C. Elliott, stattutorials.com
End