PurpuraInk Consulting- Chicago, IL & San Juan, PR

Introduction to SAS PROCedures: Data Analysis

last updated: 13SEP03

Lets PROCeed… the Saga!

As you know, SAS has two steps: DATA and PROC’s. You are probably
by now familiar with the DATA step which is discussed in detail in our
Introductory SAS seminars. You also should
refer to Introduction to SAS PROCedures for a brief
listing of the most common procedures used for statistical data analysis.

»

NOTE: These examples assume that you are using temporary data files. For
permanent data files add the library name to the file names. Also, the
appropriate <options> have been written in to get the
desired results. Other options are not mentioned here. For convenience,
SAS syntax is in capital letters, and user input in small letters. In some
cases the DATA= options has been ommitted assuming that the
PROC is being
run on the last file opened.

 

 
 
 

Data sets to practice with


 

• 
  Datasets from the book:
  A Handbook of Small Data Sets edited by D.J. Hand, et al.
• STATS@uic.edu:
  Statistics, Data Resources, and Advanced topics
• STATS@uic.edu:
  SAS II Seminar– Looking ahead and Examples

 

 
 
 

Data description and simple inference


 

1. Examine variable distributions and tests of normality

   PROC UNIVARIATE DATA=mydata NORMAL PLOT ;
    VAR v1 v2;
   RUN;
   
   Options:
   
    	
   NORMAL - runs test of normality
   
    	
   PLOT - generates stem-leaf, box, and normal probability plots. 
   
   

2. Histograms and Scatterplots
   • Histogram
     

     PROC GCHART;
      VBAR v1 v2;
     RUN;
     
     

   • Scatterplot
     

     PROC GPLOT;
      VBAR v1*v2;
     RUN;
     
     

3. Correlations

   PROC CORR DATA=mydata ;
    VAR v1 v2;
   RUN;
   
   

4. Subgroups: Overlay of plots
   • Includes setup of 2 symbols to be used
     

     SYMBOL1 V=dot; SUMBOL2 V=triangle;
     
     PROC GPLOT;
      VBAR v1*v2=v3;
     RUN;
     
     Where v3 is a categorical variable. 
     

5. Subgroups: Separate Analyses

   PROC SORT DATA=mydata; BY v3; RUN;
   
   PROC CORR DATA=mydata ;
    VAR v1 v2; BY v3;
   RUN;
   
   
   

    

    
    
    

   Multiple Regression
   

    

   • Create new output file with standardized residuals, and Cooks
     distances.

     PROC REG DATA=mydata ;
      MODEL y1=x1--x10/SELECTION=STEPWISE;
      OUTPUT OUT=mydatanew PREDICTED=rhat STUDENT=residstd COOK=cookdist;
     RUN;
     
     

   • Examination of Residuals

     PROC UNIVARIATE DATA=mydatanew NORMAL PLOT;
      VAR residstd;
     RUN;
     
     

   • Plot residuals vs. model included variables.

     PROC GPLOT DATA=mydatanew ;
      PLOT residstd * (x1 x2 x3)
      PLOT residstd * rhat;
     RUN;
     
     
      	
     where (x1 x2 x3) are the variables kept in the regression model. 
     
     

   • Plot Cook distances: first create a variable with the observation
     number to be used for the plot.

     DATA plotcook; 
       SET mydatanew;
       obsnum=_N_ ; 
     RUN;
     
     SYMBOL1=needle;
     PROC GPLOT;
      PLOT cookdist * obsnum;
     RUN;
     
     
     
     

    

    
    
    

   ANOVA- balanced design
   

    

   • Includes interaction effects and multiple comparisons of means.

     PROC ANOVA DATA=mydata ;
      CLASS x1 x2;
      MODEL y1= x1 x2 x1*x2;
      MEANS x1 x2/SCHEFFE;
     RUN;
     

   • Side by side box plots of differences between means of factors by
     their level

     PROC SORT DATA=mydata; BY x1; RUN;
     
     PROC UNIVARIATE data=mydata PLOT;
      VAR y1;
      BY x1;
     RUN;
     
     PROC UNIVARIATE data=mydata PLOT;
      VAR y1;
      BY x2;
     RUN;
     
     

    

    
    
    

   ANOVA- UNbalanced design
   

    

   • Order of categorical variable changed to get desired Type I SS.

     PROC GLM;
      CLASS x1 x2;
      MODEL y1= x1 x2 x1*x2;
     RUN;
     
     PROC GLM;
      CLASS x1 x2;
      MODEL y1= x2 x1 x1*x2;
     RUN;
     
     

    

    
    
    

   Repeated Measures
   

    
6. Syntax:

   PROC MEANS DATA=mydata  <options> ; 
    VAR <variables>
   RUN;

    

    
    
    

   MANOVA
   

    
7. Syntax:

   PROC MEANS DATA=mydata  <options> ; 
    VAR <variables>
   RUN;

    

    

   ---

   2003-9-14  VDC: WWWSTATS@uic.edu