- Contents
- Introduction
This FAQ discusses some issues you need to consider when
you conduct a survey: survey
design, data entry, and
data analysis. This FAQ is
only a starting point. Books have been written on designing
and analyzing surveys. This FAQ acquaints you with some
of the issues. Contact the
UCSS Statistical Software Consultant for additional
information.
- Back to Contents
Survey Design Tips
- When you conduct a survey, typically you get responses
like "yes" "no" and "no response". Instead of working with
these actual responses, assign numerical values to each
possible response and work with numerical data.
| Assign Values |
| Survey Answer |
Numerical Value |
| No |
1 |
| Yes |
2 |
| No Response |
9 |
- Ask questions that can be answered with a few responses
for each question and code up all possible responses (including
missing). For example, you might ask "Who did you vote for
in the last presidential election?" and make a table of
possible responses.
| All Possible Responses |
| Survey Answer |
Numerical Value |
| Bush |
1 |
| Gore |
2 |
| Nader |
3 |
| Other |
4 |
| Did not vote |
5 |
| Refused to answer |
6 |
| Missing (no response) |
9 |
- Do not ask open ended questions such as "What are the
three things you like best about UL Lafayette?" You might
get one hundred different answers from one hundred people.
- Assign a unique ID to each respondent. After you code
and enter the data, you might find that you typed the age
of a college respondent as 3 years old. If each respondent
has a unique ID, you can backtrack from the data for that
unique ID to the actual responses for that unique ID and
fix the data error.
- Suppose you will conduct a survey of 150 students. You
will record their gender and response to two questions "Did
you read a newspaper yesterday?" and "Did you watch the
news on television yesterday?" The answers can be yes or
no.
- Make a code book similar to the following for the layout
of the data file. Variable Names should
be 8 characters or less. Under the Possible Values
column put all possible values. The last two columns are
used to decide the actual layout of the data in the data
file.
| Code Book |
| Variable Name |
Possible Values |
| ID |
values 1 to 150 |
| GENDER |
1=male 2=female 9=missing |
| Q1 |
1=yes 2=no 9=missing |
| Q2 |
1=yes 2=no 9=missing |
- Prepare the code book before you conduct the survey. You
might have to modify some survey questions. Suppose you
are conducting a survey of exercise habits. Originally,
you think you should have a question "How many many hours
a month do you exercise?" and let the respondent answer
with the number of hours. You realize that this will give
you too many possible responses and restrict the responses
to "Less than five hours", "Between five and ten hours",
"Between ten and fifteen hours" and so on.
- If you are surveying a sample, be sure the sample is
random and representative of the population. For example,
if you are surveying public high school principals in Louisiana,
do not send your survey to only principals in Orleans parish.
Consult one of the books written on designing and analyzing
surveys for more information on this topic.
Back to Contents
Data Entry
You need to input the data into the computer. Your options
include:
- You can use SPSS for Windows
on UCS. When you start SPSS, you will see a Data Editor
(spreadsheet) window. Type the data into the spreadsheet,
and analyze the data using the pull-down menus.
- You can enter the data into a spreadsheet on your personal
computer and save the file as a comma delimited file. Then
you will write an SPSS or SAS
program to analyze the data on UCS.
- You can put the responses on Scantron sheets and read
the data using the Scantron machine in Stephens Hall. Contact
the UCSS Statistical
Software Consultant for more information on this option.
Back to Contents
Data Analysis
You can use SAS or SPSS to analyze your data. The
selection of one or the other is a matter of personal preference.
SAS and SPSS
are both on UCS. If you use SPSS, you can enter your data
in the Data Editor (spreadsheet) window and analyze your data
using the pull-down menus. You can also write a simple SPSS
program to read and analyze the data. If you use SAS, you
will need to write a simple SAS program to read and analyze
the data.
Some tips:
- Use variable labels and value labels (SPSS) or formats
(SAS) in your program so all output tables are completely
described. Each SPSS statement ends with a period. Each
SAS statement ends with a semicolon.
- Run frequencies of all variables first. This allows you
to validate your data. For example, if the possible responses
for gender are 1 for male, 2 for female, and 9 for missing,
then if you have a response of 6, you know you have an error
in your data entry. Possibly the data was copied incorrectly.
Fix all data entry errors and rerun frequencies.
- Examine the output of frequencies to learn more about
your data. Suppose three-fourths of your survey respondents
are male. Your survey results might be biased if you have
asked the question "Should women sports teams receive the
same amount of money as men sports teams?"
- Continue with any further analyses (such as crosstabs).
Back to Contents
Sample SPSS program for survey of
150 students
* data in comma delimited file.
* missing data is coded as 9.
title Survey of newspaper and television news.
data list file='survey.dat' free records=1
/ id gender q1-q2 .
variable labels
id 'respondent ID'
q1 'read newspaper?'
q2 'watched television?' .
value labels
gender 1='male' 2='female' 9='missing' /
q1 to q2 1='yes' 2='no' 9='missing' .
missing values gender to q2 (9) .
frequencies variables=all .
Back to Contents
Sample SAS program for survey of 150
students
/* data in comma delimited file */
/* missing data is blank */
proc format;
value labone 1='male' 2='female';
value labtwo 1='yes' 2='no';
data survey;
infile 'survey.dat' dsd missover;
input id 1-3 gender 4 q1 5 q2 6;
format gender labone. q1--q2 labtwo.;
label id = 'respondent ID'
q1 = 'read newspaper?'
q2 = 'watched television?'
proc freq;
tables id--q2;
title 'Survey of newspaper and television news';
run;
Back to Contents
Additional Information
- For additional information on any of these topics, contact
the Help Desk,
- Stephens Hall 201; 482-5516.
|
