Using SAS for Epidemiology

Charles DiMaggio, PhD

cjd11@columbia.edu

Tour of the course and materials

Why SAS?

Any computing platform 

• Unix, Linix, Windows, etc… 

Any type of data 

• ascii, delimited, Excel, etc… 
• Large data sets welcome

Any statistical procedure.  

Disadvantages 

• Yearly licensing fee
• “black box” proprietary code
• “one-size fits all” procedures
• graphics
• (bit of) a learning curve

Why Syntax?

Menu driven version available, but with syntax …

• Increased control of analyses

• Increased flexibility

• Document and recall analyses

• Create your own analyses and macros

• researchers and analysts should use the syntax-driven SAS-language

Alternatives to SAS?

You bet:

• Excel

• SPSS

• R

• EpiInfo

• Stata

• SUDAAN

1. INTRODUCING SAS

The SAS Environment

The Program Editor

Where you spend most of your time

Two flavors

• Program Editor
• Enhanced Program Editor
• Painful Lesson:  Program editor window disappears when command is submitted.
  • Use the Enhanced Editor (view - enhanced editor)

Writing SAS Statements

SAS statements start with a keyword and end with a semi-colon ;

• SAS will run statements when it sees the word RUN; 
  • proc contents data=your.data;  run;

Two ways to write comments (not run):

Tip:  Use comments liberally.

/* -------------*/  
* --------------------;

 

Submitting Statements

• Highlight by left-click dragging your mouse across the syntax 

• Click the little running icon on the tool bar

• Painful Lesson:  if you don’t highlight the syntax and just click the running icon, SAS will run every command in the Editor which may result in destroying or re-formatting data sets, etc…

DEMONSTRATION: AGE AT MI

Two Basic Types of SAS Programs

DATA steps create or manipulate SAS data sets

PROC steps conduct analyses

• SAS knows a step is beginning when it sees DATA or PROC; knows one is ending when it sees RUN; QUIT; or another DATA or PROC statement

The tyranny of the semi-colon

• In English, a semicolon is used to link two independent clauses.

• In SAS, the semicolon is used to demarcate the end of a syntax statement;  SAS likes them very, very, very much.

• Most Common Error

Interfacing With Windows

• RTF for Word documents

• “Save as…” or, highlight, R-click

• EMF for figures

• Export Utilities e.g. for Excel

Changing System Options:

• Tools - Options - System
  • e.g. center, lines per page, font type and size

Help! (with procedures)

• Help - SAS Help and Documentation - SAS Products - Base SAS - SAS Procedures - Procedures

  • command window type: “help name of procedure you need help with”
    • e.g “help means”

Help! (with functions)

• Help - SAS Help and Documentation - SAS products - Base SAS - SAS Language - Dictionary - Dictionary of Language Elements - SAS Data Set Options

  • ‘keys’ in the command window for shortcuts
    • e.g. cntrl-E to erase screen

More Help…

SAS

• Help - SAS Online Tutorial - Getting Started with SAS Software (Sample data in actual SAS environment)

Institutional

• UCLA “Starter Kit” (http://www.ats.ucla.edu/stat/sas/sk/)

Books

• ‘The Little SAS Book’ (Delwiche and Slaughter)

Google

(of course)

Coming soon

Letting go of the spreadsheet…

“Where’s my data?”

• Folks often come to program like SAS from program like Excel
• Your data is there, and you can get at it.
  • you can even bring up a spreadsheet

“Why can’t I just work with a spreadsheet like in Excel or SPSS?”

• You can, but not very useful for the large data sets for which SAS is intended.

(only) two parts to a SAS dataset

Descriptor portion

• Like meta-data (type, length, format, label)
  • PROC CONTENT

Data portion

• The numbers and characters themselves
  • PROC PRINT (or by double-clicking file in SAS Explorer)

(only) 2 kinds of SAS data

Numeric

• Missing value .
• SAS dates referenced to 1 January 1960

Character

• Missing value " "

Data set and variable names

• must start with character or underscore
• Up to 32 characters long (with no spaces or special characters like #)

2. SAS LIBRARIES

LIBNAME

LIBNAME statement creates libref

• Pointer to existing file on computer

  • libref.dataset e.g.

sparcs.mortality

must be re-issued each SAS session

• “New…”  click “Enable at startup”  name 

 Two special libraries

SASUSER

WORK

Browsing SAS libaries

• SAS Explorer

• ‘libname’ in command window

• PROC CONTENTS with ALL keyword

proc contents data=sparcs. _all_ nods;  
*/*note the space between _all_ and nods*/*
run;

INPUTTING AND READING DATA

SAS files are created in a DATA step using INPUT

• DATA – names the data set being created
  • data mylibref.mydataset; */in a named library/*
  • data mydataset; */in unnamed or work library/*
• INPUT – describes how to read the data fields
  • name the variables, identify them as character or numeric,  and specify locations

Reading data from editor window

• ‘dataline’ or ‘cards’ with semicolon on separate line

• fixed columns vs. user-specified formats (better)

• Tip: look at your data with notepad first… 

space-delimited and column input

*INPUT var1 var2 var3 \$ ;*
*INPUT var1 1-5 var2 6-10 var3 $ 16-26 ;*

• space-delimited:  INPUT list of variable names

  • numeric is default format, $ indicates a character variable

    • Caution missing values…

• column input: indicate column start and end positions

  • Can’t deal with non-standard variables, e.g.dates

reading data from external files: the INFILE statement

INFILE 'F:\\K Data\\SPARCS\\ADR02NY.TXT' LRECl=450 obs=100;

• INFILE – identifies the location of the raw data file to read

  • quotation marks e.g. “c:\temp\sparcs.txt”

  • options after the infile statement
    • LRECL= n  (each row of data is longer than the default setting, use n spaces as a row)
    • OBS= n (read only n observations, useful to troubleshoot large data set)

formatting input of external data: the INPUT statement

INPUT 
@18 DATE yymmn6.
@44 AGE 3.
@50 COUNTY $CHAR2.
;

• INPUT pointer control variable informat;
  • pointer control @n variable begins at column n
  • informats  how to read in and store data e.g. width of the field, character vs. numeric

informats

*\<\$\>informat_namew.\<d\>*

• $ indicates a character variable

• informat_name  names the informat

• w total width of the field to read. 

• . required delimiter
  • number before the dot indicates the width of the variable
    • e.g. 8.,  (number by itself without $) 8-length numeric variable

• d number of decimal points 

data documentation 

• ‘data dictionary’

  • variable names, definitions, locations, lengths, informats

  • often found on download site

  • may have to request from data vendor or source

SAS Dates

• 1 January 1960 = 1
  • December 31, 1959 = -1
  • January 3, 1961 = 368
• date informats 
  • based on the way dates are stored in the raw data, e.g. yymmn6.

behind the scenes of a DATA step 

• DATA  
  • starts process, names file
• INFILE  
  • creates temporary input buffer to hold observation
• INPUT  
  • moves and stores data from input buffer, associates attributes (e.g.length)
• RUN  
  • creates descriptor portion of file and loops around again to DATA

deciphering error messages

*_ERROR_=1 _N_=10*

• _ERROR_= binary flag, 1=error, 0=no error

• _N_=10 means happened during the 10^th loop 
  • if the data file specified correctly should be 10^th observation

• Will also get an indicator of exactly at what column SAS encountered the error, and what the error was

• SAS will be kind enough to let you know what it was expecting

common errors

ALWAYS CHECK YOUR LOG

• Worst case scenario: errors, but SAS reads in the data anyway
• Look at (some) data with text editor like notepad first
• missing semi-colon most common error

• missing RUN; next most common 

• unbalanced quotes
  • Type *’run;  
  • use * as start of a comment in case it’s not a mismatched quote
  • cancel the submitted statement 
  • click the break icon (looks like an exclamation mark), select 1. cancel submitted statements

data input tips

• SAS automatically reads in decimal points
  • only reason to specify is if they are implied but not provided in the raw data
• Don’t forget the dot after the informat
  • e.g. SAS will misinterpret $3 as a column style input of size one at the 3^rd space
• SAS assumes 256 character record (observation) length or less
  • LRECL (long record length command)
    • INFILE ‘c:\myrawdata\whatever.dat” LRECL = 2000;

some informats

• $ by itself left aligns the variable and removes leading blanks
  • $CHAR preserves leading blank values
• COMMAn. n columns of data, removes non-numeric characters like commas and dollar signs
• MMDDYY8 reads dates in the form mm/dd/yr

taming wild data sets

• slash (/) 
  • skip to and continue reading the next line of data
• #n 
  • skip to a particular line of data 
• @@ at the end of the INPUT statement
  • start a new observation regardless of whether it is at the end of a line of raw data or not

• FIRSTOBS = n

• OBS = n
  • IF _n_ = 50 THEN STOP (alternate approach)

importing Excel spreadsheets

• import wizard
  • File - Import Data - select Excel from dropdown list and follow cues…
• syntax

• PROC IMPORT DATAFILE = ‘filename’ OUT = ‘dataset’; DBMS = identifier REPLACE;

• LIBNAME 
  • libname myxls ‘c:\temp\sales.xls’;
    • SAS will treat the Excel file as a library with each spreadsheet as an individual data set