Comparing Two Independent Groups in SAS

Overview

A common situation in statistical analysis is one where you must compare two groups of subjects to determine if there is a statistical difference in the results. Depending on the situation, SAS has a number of procedures to assist in this. This section will describe some of the procedures available to you.

Introduction

In order to pick the appropriate test to use you must determine if you have independent or paired groups. Independent groups contain data on two unrelated samples. For example, comparison of male and female subjects would be independent groups. Paired groups contain a pair a measurements for each subject. The classical example is the before-and-after type of experiment where you take data before some treatment and then after the treatment and compare the two results.

Summarizing two Independent Groups

SAS provides two procedures, proc sort and proc means, that can be used to summarize two independent groups. To summary the groups using proc means the data must first be sorted. Below is an example of what they would look like:

proc sort data=dataset;
by group-variable;
run;

proce means data=dataset;
by group-variable;
var variable-list;

New users to SAS always wonder when they are required to sort the dataset before running a procedure. There is a simple way to remember, any procedure where you use the keyword by must be sorted by that variable prior to running the procedure (or in a previous step).

Example 10. Comparing two independent groups.

Often, a good preliminary way to compare groups is to graph the variable by the group variable. This can easily be done is SAS with the proc chart procedure. Below is an example of what this would look like:

proc chart data=dataset;
vbar variable-list / group=group-variable;
run;

Example 11. Charting two independent groups.

Hypothesis Tests to Compare Two Groups

Using the T-test

The T-Test is one of the classical statistical tests used in comparing groups. To use the T-Test, there are three assumptions for this test:

  1. Observations are independent
  2. observations for each group are a sample from a population with a normal distribution.
  3. variances for the two independent variables are about equal.
SAS provides the procedure proc ttest to perform simple T-Test. The format of this is:

proc ttest data=dataset;
class group-var;
var variable-list;
run;

Example 12. Using proc ttest.

Back to Sas Index Page

Author - Jack Suess
UMBC University Computing Services
Created - 1/17/96