home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
HomeWare 14
/
HOMEWARE14.bin
/
accounts
/
survey46.arj
/
SURVEY.DOC
< prev
next >
Wrap
Text File
|
1994-05-01
|
81KB
|
1,959 lines
THE SURVEY PROGRAM
CONTENTS
PAGE
SECTION 1 - GENERAL NOTES AND INSTALLATION
INTRODUCTION AND GENERAL DESCRIPTION 3
REQUIREMENTS AND INSTALLATION 5
WHAT SURVEY DOES 6
HOW THE PROGRAM OPERATES 7
DEMONSTRATION SAMPLE SURVEY 8
FILES ON THE DISTRIBUTION DISK 8
CLASSIFICATION OF QUESTIONS 9
ILLUSTRATIVE EXAMPLES 10-18
CROSS-TABULATION REPORTS 19
SECTION 2 - TUTORIAL NOTES
A TUTORIAL 22
PRINTING A REPORT TO THE PRINTER 25
PRINTING A REPORT TO THE DISK 25
PRINTING A QUESTIONNAIRE FORM 25
DESIGNING A SURVEY 26
SECTION 3 - GETTING STARTED
RUNNING SURVEY - MAKING THE PROFILE FILE 28
CLONING SIMILAR QUESTIONS 29
ENTERING RESPONSE DATA 29
CHANGING AN EXISTING SURVEY 31
1
CONTENTS - CONTINUED
SECTION 4 - SUPPLEMENTAL INFORMATION
UNUSUAL APPLICATIONS
MERGING TWO DATA FILES 32
DUPLICATING PROFILE AND DATA FILES 32
MAKING AN ADDITIONAL DEMOGRAPHIC SEPARATION 33
DELETING A RESPONSE 34
INTERPRETING THE CORRELATION REPORT 34
CHOOSING A SAMPLE SIZE 34
REFERENCES 35
2
THE SURVEY PROGRAM
SURVEY is an unpublished proprietary work
COPYRIGHT (C) 1988-1994
by Walter K. MacAdam
9 Pinewood Village, West Lebanon, NH 03784
603-643-2278
You have permission to use SURVEY, and to copy it and
distribute it to your friends and associates, or to electronic
bulletin boards, provided that the following conditions are met:
1). No fees are charged other than for distribution.
2). SURVEY may only be distributed in its original,
unmodified state,including SURVEY.EXE, MKSRVY.EXE,
RXSURVEY.EXE, SRVYADD.EXE , SURVEY.DOC. and the
demonstration files, DEMO.PRO and DEMO.DTA
You are urged to try the program on an opinion survey of your
own making. You will probably be pleasantly surprised at the
results and at the usefulness of the program.
SURVEY is not commercialized shareware. It is the original
program issued first in January 1988 and subsequently updated.
There are no stern "BUY IF YOU USE! demands. However, if you like
SURVEY, in fairness to the author, a small contribution of $10
would be a friendly acknowledgement of his efforts and would
encourage him to continue the production of useful software. Your
comments and suggestions are always welcome. Some of the features
in the current release are the result of user suggestions and are
much appreciated.
Your registration will make it possible to advise you of
future releases of SURVEY and other helpful programs.
Checks, suggestions, comments or questions can be sent to:
Walter K. MacAdam
9 Pinewood Village 603-643-2278
West Lebanon, NH 03784 walter.macadam@execnet.com
Thank you.
Standard Disclaimer
Walter K. MacAdam shall have no responsibility nor liability
to you or any other person, or persons, or entity with respect to
loss or damage caused by or alleged to be caused directly or
indirectly by use of SURVEY and/or its documentation. This
expressly includes, but is not limited, to loss or invalidation
of data, programs, files, equipment, or business opportunity or
good will.
3
THE SURVEY PROGRAM
SECTION 1 - GENERAL NOTES AND INSTALLATION
INTRODUCTION
These instructions include, in the first section, a
general description of the features and capability of the SURVEY
program. New users will find the operation much easier if this
section is printed and read before proceeding with the
step-by-step procedure described later for program operation.
GENERAL DESCRIPTION
Many organizations and individuals occasionally find it
useful to carry out surveys on a variety of subjects. Once the
responses have been collected, there comes the often burdensome
task of analyzing the results and printing them in meaningful
form. For this function, however, a computer can do much to save
time, simplify the analysis, and produce reports in the desired
format. The SURVEY program has been developed to serve this
purpose. In addition it can create and print the survey form
for distribution.
SURVEY has specific application to surveys in which responses
to questions can be identified by numbers. For these cases a
mark or number can be entered in a "box" associated with each
response choice. For questions not requiring a rank order,
respondents will be asked to place a mark in the box for the
selected item. For this reason, each box should have an
identifying number on the response form so that the marked box
number can be entered into the computer by the analyst when
inputting response data.
Questions involving rank ordering responses, or those
requesting specific quantities will have the numbers written into
the boxes by the respondents. Write-in responses should be
avoided with this system since they are difficult to handle
efficiently. In an extreme case, however, they might be
classified as a numerical type, based on the judgment of those
analyzing the results. All this suggests that caution should be
used in trying to apply this program to an existing survey
designed in another format.
4
REQUIREMENTS
The program requires a minimum of 640K of random access
memory in an IBM compatible computer. The program uses dynamic
memory allocation which allocates most of the memory requirements
at run time, based on the particular survey requirements. This
results in efficient memory use. The program can be run on one
or two floppy disks (720K or higher), or on a hard drive. If
desired, the files produced can be maintained on a disk or
directory separate from the programs.
INSTALLATION
As a first step, you should make and mark a copy of your
distribution disk. Refer to your DOS instructions for this
procedure. Always place a write protection sticker or set the
write protection slide on any original disk before copying it.
Put the original disk away for safe keeping. Installation is
simplified by the use of the included INSTALL program.
The straight-forward way to proceed is to place the
distribution disk in Drive A and type "install" (without quotes)
and follow with ENTER. Follow the directions on the screen to
complete the installation. If you are installing the program on a
hard drive, the SURVEY files will normally be placed in a new
sub-directory named SRVY. You will be given the opportunity,
however, to place them in another sub-directory of your choice.
None of your other files will be changed.
Installation on Floppy Disks
If you wish to run SURVEY on a floppy disk, it should be
one of 720K capacity or greater to accommodate the program
modules and created files. 360K disks do not have sufficient
capacity. The easiest way to transfer the files to another
formatted floppy disk is to use the DOS copy command. Place the
distribution disk in Drive A and type "copy a:*.* b:", (without
quotes), following with ENTER. The files will be copied to Drive
B. If you start with the distribution disk in the B Drive, type
"copy b:*.* a:" to place the files in A.
As an alternative, you could, of course, use the installation
program for the floppy disk transfer and type "install", following
with ENTER. At the screen asking for the destination drive, type
"B" to override the default "C".
When the files have been transferred to floppy drive B, move
to that drive and type "survey" (without quotes), following with
ENTER to start the program. This working floppy disk should not
have the write-protect notch covered, since the program will use
it to create a small configuration file, SURVEY.DEF.
5
WHAT SURVEY DOES
The program has the ability to analyze surveys involving
questions with single or multiple choice answers, and permits
straight-forward keyboard entry of all data in the sequence that
it appears in the questionnaire form, one response at a time.
The output is a detailed report and analysis, listing scores and
percent distributions in final document form ready for
reproduction. As described later, it is also possible to print a
cross-tabulation of scores and percentages for questionnaires
involving replies from different demographic groups.
At any time during the response input procedure, you can
print the current survey results to the screen or to disk as a
file in ASCII format. You could also print these intermediate
analyses to the printer, but you can save time and printer paper
by waiting until all pages in the screen report meet your final
requirements and an appropriate number of replies has been
processed.
Input can be temporarily discontinued at any time, with the
response data stored on the disk until the work is resumed. For
each questionnaire, up to 100 questions can be accommodated, and
most may have up to 40 choices. More than 1000 responses can be
included, although most surveys can obtain results providing a
reasonable degree of confidence with only a few hundred. The
practical limitation is the amount of random access memory. Any
number of surveys can be handled, with separate profile and data
files kept on the disk or directory.
In addition to being able to summarize and analyze the total
of all responses, the program has the capability of summarizing
results by a demographic subdivision. For example, separate
analyses can be made in terms of a single classification such as
one by age group, gender, or the geographical area in which the
respondent is located. The program accommodates one such
subdivision. If this demographic classification is desired, all
that is necessary is to indicate at the beginning of the set-up
procedure that a demographic separation is required, and to provide
the number and titles of the desired groupings or "populations".
The printed analysis following data entry will then include
summaries for: (1), all responses; (2), responses from each
classification; and (3), responses from those who failed to make a
demographic selection. Up to nine demographic groups can be
accommodated in this single separation.
The final printed results are produced in report form, ready
for reproduction without retyping or re-formatting. The report
can also be generated as a file in ASCII format suitable for
editing with a word processor. At any time, a print can be made
to the screen to inspect progress. The report can be produced in
a continuous print without page breaks, or can be printed in
"form feed" or on separate sheets with a pause between sheets
6
for paper replacement. In the latter two cases, a feature of the
program prevents splitting a response summary for a single
question between two printed pages. Because of the program's
ability to produce a comprehensive and detailed report, it is
advisable to consider, at an early stage, the number of printed
pages that would result, based on the number of planned questions,
choices and demographic groups. This aspect is discussed later in
in these instructions.
An important capability of the program is the option to
examine and report a possible correlation between the responses to
two different questions. For example, a commercial organization
might want to test the effectiveness of an advertising program on
a local radio station. This might be done by noting the response
to a brand recognition question by those who indicated, in
responding to a separate question, that they listened to the
station, and then comparing this with this recognition by those
who did not. The program can include the correlation in terms of
a percentage figure and can include the results in easily
understandable format in the printed results. This makes it
unnecessary to try to discern possible correlation by developing
demographic groupings for answers to a single question, and
trying to probe the voluminous data for possible correlation.
An additional feature for surveys involving several
demographic groups permits printing the results for a question
as a "cross-tabulation". This displays a comparison of scores
and percentages for question response choices, cross tabulated by
demographic group. For example, one can read the percent of the
respondents selecting a particular choice, as well as any group's
contribution to the total selection of that choice, expressed as
a percent.
HOW THE PROGRAM OPERATES
The SURVEY program first asks for some information on the
format of the questionnaire and stores this on the disk as a
profile file for later use. As covered in more detail in a later
section, this information includes the type of questions, the
number of demographic groupings, as well as the number of choices
in the answer to each question. If correlation measurement is
required, this information is requested. The report title and
desired text for use in the summarized output report is also
stored on the disk in this one-time operation.
With the profile completed, the program will print a sample
survey form to the disk as a text file in ASCII format with
the extension ".FRM". In addition, the user can print this
sample questionnaire form to the printer at this point if
desired. Following creation of the profile file and survey
form, the user is returned to the main menu to select options
for entering the response data, one response at a time.
7
The response data entry process permits correction of
errors and re-entry of data. If necessary, the response to the
previous question can be re-entered. The information is entered
from the keyboard in the same sequence as it appears on the
response form. The program prompts for choice selection numbers
for each question. The summarizing printed report includes the
score for the response choice in each question and a percentage
valuation for each of the choices. In addition, for each
question, there is a count of the cases in which there was no
entry by the respondent.
DEMONSTRATION SAMPLE SURVEY
To provide an opportunity to become familiar with program
operation and data entry, a short sample survey file, DEMO.DTA
and its profile file, DEMO.PRO, are included in the distribution
disk. You can access these files and print the results of this
sample survey on the screen. You can also gain experience in
data entry by adding more response data of your own choosing
and noting the changed results. The DEMO questions are the same
as several of those included on example forms later in these
instructions.
FILES ON THE DISTRIBUTION DISK
The programs on the distribution disk consist of those shown
below. At least all but SURVEY.DOC, READ.ME, and INSTALL.EXE
are needed in the drive/directory used to run SURVEY and use the
demonstration program. INSTALL.EXE is the installation program.
README is a short informational file that you can read by typing
"type readme" and folllowing with ENTER.
SURVEY.EXE, MKSRVY.EXE, RXSURVEY.EXE, SRVYADD.EXE,
DEMO.PRO, DEMO.DTA, SURVEY.DOC, READ.ME and INSTALL.EXE.
If you should want to send the SURVEY program files in
compressed form to a computer bulletin board, all the listed
files must be included. They should be compressed in ZIP
format, with the compressed ZIP file given the name SURVEY46.ZIP,
assuming you are sending Release 4.6. If a higher release number
is involved, the name should be changed to correspond.
When you create a profile for your own survey, the program
will create a format file with a name you have chosen, with the
extension, "PRO". Response data for this survey will be kept in
a file with the same root name, but with the extension "DTA".
These files will have titles assigned by the user and will be
located on the drive/directory assigned in the setup procedure.
An additional file, SURVEY.DEF is created and used to determine
8
whether the opening menu can be by-passed, and to keep a record
of any default directory for files. For this reason, if you
are running the program on a floppy disk you should not add a
write-protect tab or set the write-protect slide.
CLASSIFICATION OF QUESTIONS
Since survey questions vary in such matters as number of
choices available and may offer the possibility of rank ordering
the selections, the program will need to know in advance the type
of question involved. It will also need to know the number of
choices available and the title or designation for each choice.
This information is stored in the profile file.
Most survey questions can be classified into one of four
general types which we designate as follows:
TYPE A Questions with two or more possible response
choices, but with only ONE CHOICE permitted.
Typical choices might be based on opinions,
subject categories, votes or numerical ranges.
The choice is made by placing a mark in a "box".
For this type, the respondent does not enter a
number.
TYPE B Questions with two or more possible choices with
NO LIMITATION on the number of choices that
might be made. Respondents mark their selected
"boxes" and do not enter numbers.
TYPE C Questions in which the respondent is asked to
RANK ORDER the preference for each choice. The
respondents enter numbers in rank order.
TYPE D Questions in which totals are progressively
accumulated for each response option as each
response is processed. The final report will
total the number entries for each selection
option.
Questions making a demographic separation are, of course,
single-choice questions, and the program will automatically
classify them as such.
Only one demographic separation can be included, and its
possible choices must be mutually exclusive. A respondent can
meet only one of the category classifications.
9
SOME ILLUSTRATIVE EXAMPLES
Before proceeding with detailed instructions on procedure, it
may be helpful to see examples of the different types of
questions that can be handled by SURVEY, how almost any type of
multiple choice questionnaire can be analyzed, and how the
results are included in finalized report form. The tables on the
following pages present examples of several types of multiple
choice questions and the typical response summaries that are
derived by the SURVEY program.
If your survey requires a demographic subdivision, remember
that ANY DEMOGRAPHIC SEPARATING QUESTION MUST ALWAYS BE THE
FIRST. Note that, in this case when entering response data, the
program will automatically prompt you for the answer to this
question as Question No. 1.
Table I is a sample of a question making a demographic
separation between Northern and Southern company divisions.
Remember that DEMOGRAPHIC SEPARATIONS MUST BE CHOSEN SO AS TO BE
MUTUALLY EXCLUSIVE. In other words, an individual respondent can
only fit into one of the categories. For example, you can NOT ask
for both age group and gender in the same question or have
more than one demographic separating question. This
limitation can be largely overcome, however, by using the
cross-tabulation feature or combining data files as described
later.
Note that the optional responses are numbered. This is required
to identify the selected choice when the data is entered during
program operation. Note also that double quotation marks can not
be included in any question text or title. These are used for
dividers in the profile file.
- - - - - - - - - - - - - - -
TABLE I
SAMPLE DEMOGRAPHIC QUESTION
Q1. Please place a mark in the square corresponding to the
regional division in which your company is located.
__
1. |__| Northern
__
2. |__| Southern
10
Table II illustrates a situation in which three types of
questions are included. The first type requires only one answer
and, for the purposes of the program, is designated Type A. The
second permits one or more choices, listed in any order, and
without order of preference. This is designated Type B. The
third question on Table II, Type C, permits more than one choice,
but respondents are asked for all answers in the descending order
of preference, 1 being the highest in rank. Note that the
question must be numbered as well as the possible choices for a
reply to each question.
11
TABLE II
SAMPLE SURVEY
(Abbreviated)
Q2. Please mark the square corresponding to the number of
employees in your company. Mark only ONE.
__
1. |__| Less than 10
__
2. |__| 10 to 25
__
3. |__| 26 to 50
__
4. |__| 51 to 100
__
5. |__| Over 100
Q3. Please mark ANY of the following factors having an IMPORTANT
positive influence on your company's profits.
__
1. |__| Good Labor Market
__
2. |__| Climate - Environment
__
3. |__| Nearby major airport
__
4. |__| Nearby rail facilities
Q4. What are the most IMPORTANT reasons for your selecting
particular brands of computer software? Please rank 1,2,3, etc.
for the following reasons in descending order of importance.
__
|__| Manufacturer's reputation
__
|__| Price
__
|__| Dealer support
__
|__| Recommendation by associates
12
Table III illustrates the typical analysis print-out
produced by SURVEY when processing survey responses of the type
shown on Table II. Note that both a numerical score and a
percentage valuation is printed for each response choice, and
that a count is made of those who failed to make a response. In
each case the results are sorted in descending rank order so as
to provide easy recognition of the most important choices.
"Score" is the count of the respondents who have made this
particular choice.
13
TABLE III
SAMPLE SURVEY RESULTS SUMMARY
PRINTED BY THE SURVEY PROGRAM
There were 10 responses entered from this questionnaire.
QUESTION 2. (Single Choice)
For responding companies, the employee numbers ranged as follows:
CHOICE SCORE PERCENT
10 to 25 3 37.5
Over 100 3 37.5
Less than 10 1 12.5
26 to 50 1 12.5
51 to 100 0 0.0
Made no response 2
QUESTION 3. (4 choices maximum, without ranking)
Those responding marked important factors influencing profits as:
CHOICE SCORE PERCENT
Good labor market 6 37.5
Nearby rail facilities 5 31.3
Climate-Environment 3 18.8
Nearby major airport 2 12.5
Made no response 1
QUESTION 4. (4 choices maximum, with rank ordering)
The most important reasons for selecting software were ranked:
CHOICE SCORE PERCENT
Manufacturer's reputation 32 32.0
Price 25 25.0
Dealer support 25 25.0
Recommended by associates 18 18.0
Made no response 0
14
Table IV illustrates a questionnaire format frequently
used to assess gradations in views on certain questions or
statements. In this example, only ONE ANSWER is required to each
question. It is therefore of the type designated "A", as previously
noted. The gradations in the respondent's view in this case are
"Agree Strongly", "Agree", "Undecided", "Disagree", and
"Disagree Strongly", and numbered 1 to 5.
The Table IV format can easily be handled by the SURVEY
program by entering the column numbers for the individual
choices. In this case, the results will be printed out in the
format of Question 2 in Table III. The choice titles would then
be "Agree Strongly", "Agree", "Not Decided", etc. and would be
numbered 1 to 5 in this example. The scores and percentages would
be indicated for each choice.
Note that there are four questions in Table IV. Assume, for
example, that a respondent placed a mark in the "Strongly Agree"
box in the first question, Q5. When you are posting this response
using the SURVEY program, and you are asked for the choice number
selected in Question 5, you would type a "1", since the marked
box was in column 1. Only one box should have been marked. This
is a single-choice question, one that we define as Type A.
The same format can be used for questions such as:
"On a scale of 1 to 10, enter your opionion on the
following statements"
This would require 10 numbered columns instead of the 5 in the
example in Table IV, but would be processed in the same manner.
If there are several successive questions of the same type in a
survey, with the same choice titles, a feature of the program
permits one to "clone" a following question format when
establishing a the profile file. Then only the question and
response summary text need be entered. This saves considerable
time and effort.
15
TABLE IV
SAMPLE SURVEY
(Abbreviated)
Please mark the box for your answer the following questions.
1 2 3 4 5
AGREE NOT DISAGREE
STRONGLY AGREE DECIDED DISAGREE STRONGLY
__ __ __ __ __
Q5. Do you believe that |__| |__| |__| |__| |__|
Television has had a
bad influence on
children's morals?
__ __ __ __ __
Q6. Do you believe that |__| |__| |__| |__| |__|
air bags should be
required in all
automobiles?
__ __ __ __ __
Q7. Do you believe that |__| |__| |__| |__| |__|
the U.S.should work
toward a balanced
budget?
__ __ __ __ __
Q8. Are you willing to |__| |__| |__| |__| |__|
give up your frost-
free refrigerator
to reduce energy
consumption?
16
Table V illustrates a simplified example of a printed
analysis report of the results of a graded opinion question of
the type displayed in Table IV. Since only one entry is
acceptable, this is classed as a Type A question, even though the
format is somewhat different from that shown in Table II as
Question No.2.
17
TABLE V
SAMPLE SUVEY RESULTS SUMMARY
PRINTED BY THE SURVEY PROGRAM
(Abbreviated)
QUESTION 5. (Single Choice)
Belief that TV has a bad influence on childrens' morals varied:
CHOICE SCORE PERCENT
Agree Strongly 4 40.0
Not Decided 3 30.0
Agree 1 10.0
Disagree 1 10.0
Disagree Strongly 1 10.0
Made no response 0
18
CROSS-TABULATION REPORTS
When a survey includes several demographic separations,
it is sometimes helpful to present a view of the choice scores
for the individual demographic groups in a format suitable for
comparison. This can be done by constructing a matrix or cross-
tabulation, with choice selection results for each group listed
vertically, and group preferences for each choice listed
horizontally. The question being analyzed in this fashion
should be either of the Types A or B described earlier. A
cross-tabulation report can be sent to the printer for questions
of these types by selecting the PRINT CROSS-TAB Option in the
VIEW/PRINT column of the Main Menu.
Since a cross-tabulation may be a little confusing at
first sight in the case of a typical multi-choice question
involving several demographic groups and many responses, it is
best to look at an over-simplified set of survey results such as
produced by the short DEMO survey file. This will illustrate the
general format and permit comprehension of the way in which
percentages are derived. The DEMO file as present on the
distribution disk includes only ten fictitious responses and
only two demographic groups. In this very simple example,
Question 2 has been selected for the print of a cross-tabulation,
with the results shown in Table VI.
19
Referring to Table VI, it can be seen that demographic
group results are listed in vertical columns, first for the
total, and then for each demographic group. The total scores for
each group are listed on the first line with the left margin
heading "ALL RESPONSES". The third line shows the percent
contribution of each group to the total response count, eight in
this case. The 6 responses in the Northern group contributed
75% of the total of eight. Dropping down to the next line, we see
an analysis for the first choice, "LESS THAN 10" employees.
There was one selection for this choice, and this was made by a
respondent in the Northern group. Since the total Northern group
score was 6, this single score is 17 percent of the group total.
This is shown on the second line of the "LESS THAN 10" choice
block. Note, however, that this selection represented one of the
8 total responses. The third line titled "% of Total" shows this
therefore to be 13%. Proceeding in this fashion through the
remainder of the table, one can see:
The contribution of each Group Choice to the Group Total
The contribution of each Group to each Choice Total
The contribution of each Group Choice to total of responses
While the example in Table VI is over simplified, a cross
tabulation of this type will be found particularly useful for
surveys with several demographic groups. Note that printing a
cross-tabulation for a larger number of demographic separations
will in some cases require the use of a wide carriage printer, or
compressed print (12 or 15 cpi) on a standard narrow width
printer. The program will warn the user if this is the case.
20
TABLE VI
CROSS-TABULATION OF RESPONSE CHOICES VERSUS DEMOGRAPHIC GROUPS
For QUESTION NUMBER 2
For the responding companies, the employee numbers ranged:
CHOICE TOTAL NORTHERN SOUTHERN
ALL RESPONSES 8 6 2
% of Group 100% 100%
% of Total 100% 75% 25%
LESS THAN 10
Choice Count 1 1 0
% of Group 17% 0%
% This Choice 100% 100% 0%
% of Total 13% 13% 0%
10 to 25
Choice Count 3 3 0
% of Group 50% 0%
% This Choice 100% 100% 0%
% of Total 38% 38% 0%
26 to 50
Choice Count 1 1 0
% of Group 17% 0%
% This Choice 100% 100% 0%
% of Total 13% 13% 0%
51 to 100
Choice Count 0 0 0
% of Group 0% 0%
% This Choice 0% 0% 0%
% of Total 0% 0% 0%
OVER 100
Count Choice 3 1 2
% of Group 17% 100%
% This Choice 100% 33% 67%
% of Total 38% 13% 25%
21
SECTION 2 - TUTORIAL NOTES
A TUTORIAL -Viewing and Entering Data in the DEMO survey
To become familiar with the data entry procedure and the
several program features that ease this somewhat boring activity,
the first-time user should first access the short simplified
demonstration survey files, DEMO.PRO and DEMO.DTA, included on
the distribution disk. This is the survey used in the
illustrative examples in the preceding tables. DEMO.PRO is the
profile file and DEMO.DTA the data file for this sample survey.
The data file includes only a few response entries and provides a
good opportunity for a first-time user to add one or more
fictitious responses to each of the questions and then print the
results to the screen to see the changes.
When entering data for a typical survey with several
questions and a relatively large number of responses, there is
some advantage in having the operation performed by two people.
One can read the response sheet, announcing the question number
and the numbers the respondent has entered into the individual
"boxes", while the other keyboards this data into the computer.
In the case of the short DEMO sample survey this method is
probably unnecessary. The procedure is as follows:
Viewing an Existing Survey Report
Before entering any new data it will be interesting to see
a screen print of DEMO as it currently stands. When you start
working with your own surveys you will find these screen displays
very useful in checking your report format and evaluating progress
as data entry proceeds.
The first step is to start the program and identify the
profile and data files desired and the directory in which they
can be found. Move to the directory holding the SURVEY program
and type "survey" (without quotes). You will see the title page.
Press any key to move to the Main Menu. Since the DEMO files are
already established, use the arrow keys to move the highlight bar
to the VIEW/PRINT column and press ENTER. The VIEW A REPORT line
will be highlighted initially. Press ENTER. (Instead of using
the arrow keys, you could have simultaneously pressed the ALT key
and the highlighted "Hot Key" for this window, namely, ALT-V.)
You will next be asked to identify the directory holding
the files you wish to access. Since DEMO is in the same
directory as the program files, just press ENTER.
With the directory chosen, you will see a directory screen
listing the available data files in this directory. If you are
just starting you will see only one, "demo.dta". type "demo"
here. Follow with ENTER. (The extension, ".dta" need not be typed).
22
You will then be asked whether the question choices should
be displayed as sorted in descending order of response scores, or
listed as they appeared on the questionnaire form. If you would
like to see them sorted, press "1." You will next see the first
page of the report. This indicates that there are two
demographic groups in this survey, "Northern Companies" and
"Southern Companies." Moving to the next screen by pressing a
key, you will see a report of the scores and percent distribution
of the entries made thus far for this demographic question Number
1. Press a key to move to similar analyses for the remaining
questions. The report at this point is summarizing the combined
scores of both demographic groups. You can then continue to view
the results for each of the demographic groups separately and
ultimately return to the Main Menu.
Adding a Response to an Existing Survey
Now see what happens when you add a response of your own
to this demonstration survey. You can do no harm to the program.
The file is there for you to play with. Since you are adding
data to an existing survey, use the arrow keys to move to the
DATA IN column and press ENTER or down arrow. Press the down
arrow to highlight the second line heading, MORE DATA ADDED and
press ENTER.
You will again be asked to identify the file name to
access. Select DEMO again. The questions you will be responding
to are the same ones included in Tables I, II and IV.
To simplify your entry, if you wish, you can take a
pencil and mark the "boxes" in Tables in these instructions to
correspond to your arbitrary new choices. Just follow the
directions on the screen as you proceed. Note that you will only
see the question numbers on the screen and not the questions
themselves. The Response Number of the current response appears
in the upper right of the screen.
The first question establishes the demographic group to
which the respondent belongs. Type either "1" or "2" since there
are only two such groups. Follow with ENTER. Since there is only
one choice, you will immediately be moved to the data entry
screen for the second question which is also a single choice, or
Type A question. It asks for the range into which the employee
numbers of the respondent's company falls. Type any number
between 1 and 5, inclusive. Pressing ENTER following this number
will move you to the next screen.
23
Question 3 is a multi-choice question and, as indicated
in the directions on the screen, you can select any or all of the
available choices. This makes it a Type B question according to
our classification. This question corresponds to Question 3 in
Table II. Select any or all of the maximum of four choice
numbers, following each with ENTER. If you select less than
four, press X after the last selection to let the computer know
you are through with this question. If you select all four
choices or press X you will see the screen for Question 4.
Goof Protection
If you are in the process of entering data for a multi-
choice question such as No 3 and make a mistake, you can re-enter
all the data for the SAME question by pressing ESC. The entries
that you already made for that question will be disregarded and
you can start the question response again.
Question 4 is a rank order question, or Type C. It
requires entering a rank order number in each of four boxes, with
the highest ranking 1, progressing in descending order to 4.
Type any sequence, 1,4,3,2, for example. Follow each with ENTER.
Super Goof Protection
After you have successfully completed answering Question
4 and are viewing the data input screen for Question 5, you are
now in a good position to try out another one of the very useful
features of SURVEY. This is the ability to go back and repeat the
entry of data for the PREVIOUS question. All you have to do is
press Page-Up and you will see the data entry screen for the
previous question. Then every entry you have made beyond the
beginning of that question will have been deleted. Therefore you
can re-enter the data correctly if you think you may have made an
error. You will see the question number flashing, indicating
that you are re-entering a question response. You will not be
able to go back, of course, when you are seeing the entry screen
for Question 1.
Question 5 in the DEMO file is a single-choice graded
opinion question corresponding to Question 5 in Table IV. It
asks if the respondent believes that TV has had a bad influence
on childrens' morals. Enter any number 1 to 5 corresponding to
the column numbers headed "Agree Strongly", "Agree", etc. Your
entry will be recorded and you will see the Question 6 display.
Question 6 in the DEMO sample questionnaire is the type
that accumulates response numbers entered in each choice "box" as
the responses are processed. This is a Type D question as
previously described. It can be used for accumulating such items
as votes for different candidates or inventory of supplies etc.
Just enter any number (0 or multi digit=ok) for each of the three
boxes, following with ENTER, or press ENTER alone if you want to
skip a box and consider it to hold a zero.
24
Question 6 is the last question. When you complete it,
you will be asked if you want to enter data for another response
sheet or return to the Menu. For the purposes of our trial run,
type "N" and return to the Main Menu. By typing "X" instead
you could have repeated the entire input of all questions for the
last response if you felt that an error still existed. Pressing
ESC here would save all the entries made thus far, and then exit
the program.
Now, with the Main Menu in view, you should select the
VIEW/PRINT column and choose VIEW A REPORT to see a report of the
DEMO survey results which now include the data entries you have
made. The number of responses will have increased by one and the
scores and percentages will have changed based on your entries.
If you continue the display to include the two demographic groups,
you will see that the changes have taken place only in the group
you selected in the demographic Question 1.
PRINTING A REPORT TO THE PRINTER
If you wish to print the report to the printer, you would
select VIEW/PRINT - PRINT A REPORT in the main menu and proceed to
select the appropriate data file. Turn on the printer and follow
the directions on the screen.
PRINTING A REPORT TO THE DISK
The report can be sent to the disk as a text file in ASCII
format suitable for editing with a word processor if desired. This
is done by selecting VIEW/PRINT - SEND REPORT TO DISK in the main
menu. Select the appropriate data file when prompted. The text
file will have the root name with the extension ".RPT."
PRINTING A QUESTIONNAIRE FORM
You can print out a copy of the questionnaire form for the
DEMO survey example. To do this, move to the VIEW/PRINT column in
the Main menu and select MAKE SURVEY FORM. Select "demo.pro"
from the directory list and have your printer turned on. Follow
directions to print the questionnaire. The program will have
previously created this as a text file on the disk with the
extension ".FRM."
25
DESIGNING A SURVEY
Designing a useful survey is not an easy undertaking. Much
thought must be put into selecting the appropriate questions and
response choices, and limiting their number to reasonable values.
No attempt will be made in these instructions to provide a
comprehensive discussion of survey design. Several reference
texts are listed at the end of these instructions in the event
that users want to dig deeper into the subject. To make the
SURVEY program fully effective, however, the following guidelines
should be observed.
1. Make sure your survey consists of sequentially numbered
questions and that each optional response is also designated by a
number. These numbers are for your use in entering response data
and have no significance as far as the respondents are concerned.
When responses are received and before entering data, you may
find it convenient to number the sheets sequentially. This is to
help you avoid entering a response more than once and to
correlate with the response numbers referred to in the SURVEY
program.
2. If you try to adapt this program to the responses from an
existing questionnaire produced without reference to this
program, and in which the demographic information is not obtained
in the FIRST QUESTION, your data input will be seriously
affected. Design your surveys with the program in mind and
produce the profile file in the normal manner. Remember that you
can not use double quotation marks in any question text or title
this file.
3. If, after hearing the warnings to place the demographic
question as the first question, you still believe it important to
have the demographic question last in your survey, (a question
on gender, possibly) you MUST still enter it first in SURVEY when
processing the responses. In this exceptional case you MUST
number the questions in the questionnaire sheets starting with
Number 2 to correspond to the question numbers identified on the
screen as the responses are processed. This complicates data
entry and is a possible source of error. Avoid this procedure if
possible. All DEMOGRAPHIC SUBDIVISIONS MUST BE MUTUALLY EXCLUSIVE
with only a single choice possible..
4. Avoid write-in response options if at all possible. They
are difficult to classify. if necessary, collect these separately.
5. Keep in mind, that for the purpose of printing the final
report, you will be asked to provide a "Response Summary
Statement" for each question. Example: "For the responding
companies, the employee numbers ranged as follows:". This
statement should not exceed one line in length. (You could
substitute a one-line question text here if you prefer).
26
6. Make sure you can classify each question into one of the
following types:
A. Only ONE selection can be made by marking a box
B. Any or all response options can be selected by
marking boxes
C. Selection of options in descending rank order by
entering numbers
D. Numbers entered in any option box are cumulated in
final report.
7. The number of printed pages produced in the final report
should be carefully considered in the design process. You should
assume that no more than an average of about four questions can
be analyzed per printed page. If you have demographic
subdivisions this will multiply the number of pages by one more
than the number of demographic separations. In the interest of
maximizing the survey returns and making a useful report you will
want to limit the number of questions and question choices to
those providing the most meaningful and useful information. As an
example, an average 40-question report with six demographic
subdivisions might require about 70 pages. As mentioned earlier,
you should use the screen print option for most of your
intermediate viewing of the report results.
8. Before distributing the questionnaire you may find it
helpful to run the SURVEY program, make your profile file, move
to the DATA IN column in the Main Menu, select ADD DATA 1ST TIME,
choose your data file name, and then enter two or three fictitious
responses. You should then use VIEW/PRINT - VIEW A REPORT Option
to view the preliminary results on the screen. At this time, do
not send any output to the printer. Check the screen display to
see that the desired format is achieved.
When you are ready to start entering actual responses, be
sure to first select the DATA IN - ADD DATA 1ST TIME Option. It
will delete any of the preliminary trial data in your data file.
If you later temporarily discontinue genuine data entry, remember
to use the MORE DATA ADDED Option when you start adding data to
this file in another session. Don't use the ADD DATA 1ST TIME
Option again unless you want to start over from the beginning!
27
SECTION 3 - GETTING STARTED
RUNNING SURVEY - MAKING THE PROFILE FILE
Before running SURVEY for the purpose of establishing a
profile file, you should have before you a sample response form
for the questionnaire you are proposing to distribute. This will
help in answering the questions which will be asked in the
process. This can be in draft form because the program will offer
you the opportunity to print out the questionnaire form
when you complete the entries for the profile file. It will also
automatically print this form to the disk for possible editing
with a word processor.
With the DOS prompt showing in the drive/directory holding
the SURVEY programs, start the program by typing "survey"
(without quotes). Follow by pressing ENTER. You will see the
introductory screen and can press any key to go to the Main Menu.
If you are a first-time user, you should press ENTER at the
LEARN/HELP column of the Main Menu. Selections in the drop-down
window will provide information on the use of the program and
offer several examples. You will ultimately be returned to the
Main Menu. Note the changing "Explanation" at mid-screen as you
move from one window element to another.
You should read the Tutorial included earlier in these
instructions. At his point, you may also want to move to the
VIEW/PRINT column and select VIEW A REPORT to obtain a screen view
of the short demonstration survey, DEMO. You will see a list of
the available data files in the directory and can select DEMO.DTA
for viewing. When satisfied, you can return to the Main Menu.
You will next need to establish the profile file for your
survey. To do this, use the arrow keys to move to the CREATE
column or type the hot key, "C". Select MAKE NEW SURVEY in the
drop-down window. Since this is a new survey, you will be asked
to select a directory to hold the profile and data files. You
will also need to create a file name for the profile file and for
storing the survey data. This must be no longer than eight
characters, without any extension. This file name will be used
for the profile file with the extension ".PRO". The file to hold
the data will have the same name with the extension ".DTA".
You will be asked for the number and names of any demographic
groups, the text of questions, and the number and names of each
question choice. One demographic question is permitted and up to
nine demographic separations can be made within this question.
Remember, however that the length of the report is multiplied by
one more than the number of demographic groups. A statement
will also be needed for each question to be used as a question
heading in the report. Establishing the profile file is a one-
time operation for each survey. It can be used for future
28
surveys with the same characteristics. You can not use double
quotation marks in any of your entries. These are used by the
program to separate entries in the file. You have the option of
analyzing the correlation, if any, between the answers to two
different questions. If you wish to include such a correlation
analysis, you will be asked for the question numbers and choices
involved. Be careful in making the profile file. Watch for typos.
CLONING SIMILAR QUESTIONS
If your survey includes successive questions such as shown in
TABLE IV, in which the choice titles are repeated (Agree Strongly,
Agree ..., etc., for example), you can save time and effort by
using the "clone" option in creating the profile file. When asked
if you have another question to add, you can type "C" instead of
"Y", and the next question will be cloned. The question type,
number of choices and choice names will remain unchanged. You
will then only be asked for the new question and response summary
text for the added question.
PRINTING THE QUESTIONNAIRE FORM
When all the set-up questions have been answered, the program
will automatically create the questionnaire form as a text file in
the current directory or diskette. You will then be asked if you
wish to print a Questionnaire Form to the printer based on the
information you have entered in the profile file. This is a
useful form to have, at least for record purposes. If you want to
create a fancier form later with your favorite worde processor,
you can use this as a model. If you want to print this form later
you can still do this by moving to the VIEW/PRINT column and
choosing the MAKE SURVEY FORM option in the VIEW/PRINT drop-down
window.
ENTERING RESPONSE DATA
On completion you will be returned to the Main Menu. If you are
ready to enter data from some or all of the responses to your
questionnaire, or experiment with some test entries, you can select
the DATA IN column in the Main Menu and choose ADD DATA 1ST TIME
Follow the instructions for adding the data.
In the ADD DATA 1ST TIME Option, you have the choice of
entering some fictitious test response data as previously
described in Section 8 of "DESIGNING A SURVEY", or you could
proceed with entering data from a genuine response. You will see a
list of the data files in your directory. Type the file name for
storing the survey data and follow with ENTER. Be sure to enter
the SAME NAME as the one you used in setting up the survey profile.
You will remember that the profile file had the suffix ".PRO". The
data file will have the same name, but with the extension "DTA".
(In case the new data file is not listed, type it anyway in case it
was deleted by accident).
29
With the file identified, you will be asked to enter the
responses for each question, taking one response sheet at a time.
If you are entering test data, remember to re-select ADD DATA
1ST TIME before entering data from genuine responses in order to
erase the test data by over-writing it.
When prompted on the screen for data entry from a question
response, you will only see the Question Number and Choice
Number, not the question text. If you are entering data from a
multi-choice question, you must follow any number you type with
ENTER.
In the case of Type B questions which permit any number of
choices without ranking, enter only those numbers selected. When
all the selected Type B choices have been entered, type the
terminating "X". (For consistency press ENTER before typing "X",
but if you type "X" immediately after the last entry, that entry
will still be recorded).
Type C and D questions require you to make an entry for EACH
possible choice, whether or not the choice has been selected.
Follow each choice number with ENTER, or type ENTER alone if a
choice is vacant. Fear not; the appropriate directions are
included on the screen as you process each question.
The directions should be easy to follow. If you make an
error in entering data in a multiple choice question, all is not
lost. You can re-start the entries for this question in this
response by pressing ESC. In addition, if you think you have made
an error in entering data in a particular response but have gone
by the question involved, you still have the opportunity of re-
entering data for the previous question by pressing Page-Up. You
can also repeat entry for all the questions in the most recent
response when you come to the end of the last question on the
response sheet. At that point, you will also be asked if you
wish to continue entering data for the time being. By typing "N"
you will return to the Main Menu, and the responses data to date
will be stored on your disk. Here you can choose VIEW A REPORT in
the VIEW/PRINT column to print the report to the screen as it
currently stands. If you want to take a break, you can exit to
DOS from the Main Menu by moving to QUIT. Your data will be
stored, ready for your next session. DO NOT TURN OFF YOUR COMPUTER
UNTIL YOU HAVE EXITED THE PROGRAM, OR YOUR DATA MAY BE LOST.
You will be warned if you try to use ADD DATA 1ST TIME for an
established survey, since this creates a new data file. If you
use a file name with this option for a valuable file that already
exists, this might wipe out your data. The procedure for adding
responses to an existing data file requires the selection of the
ADD MORE DATA choice. Do not use ADD DATA 1ST TIME here. The
procedure is otherwise the same as in the case previously
described for starting data entry for a new survey.
30
CHANGING AN EXISTING SURVEY
In some instances, after a profile has already been
created, you might need to make editing changes in the text of
the Report Title, a Response Summary Statement, or to modify a
Choice Title in a question in an EXISTING survey. Alternatively,
you might want to add or delete a question in an existing
questionnaire and start over with a new survey. You can even add
or delete a question choice. Instead of repeating the detailed
effort of creating a completely new profile file to accommodate
the change, you could use one of the eight choices in the
drop-down window of the REVISE column. These editing options
include:
■ Options to edit either the Report Title, a Question
Text, Response Summary Statement, or Choice Title of
a question in the existing survey profile file.
■ An option to add one or more questions to an existing
survey (Resulting in a new survey profile for a new
survey).
■ An option to delete a question from an existing survey.
(Resulting in a new survey profile file for a new
survey).
■ Options to add or delete one of the response choices
for a particular question.
The first option group could be used, for example, if you
decided that the text in one or more of the question texts,
response summary statements, or the choice title could be
improved or corrected. No change in the survey questionnaire
would be required since the changes only affect the display of
the survey results.
The remaining options make significant changes to
questions in a survey. Under these conditions new survey profile
and data files will be created. They will have a new root name
unless you choose otherwise. Because of these major changes, any
data file produced in the original survey can not be used with
the new profile or data file. In other words, you would have to
start entering data with a new or revised questionnaire.
These last options have application where you want to
start a new survey very similar to one used previously, and you
want to avoid the trouble of creating a completely new profile
file. By using a new name, you would not change the original
profile and data file. If you selected the same name as a
previously created survey analysis, the former profile and data
file would be written over and would be no longer available.
31
SECTION 4 - SUPPLEMENTAL INFORMATION
UNUSUAL APPLICATIONS
MERGING TWO DATA FILES
On occasion, there may be the need to merge two data
files created by SURVEY, using the same profile file and
questionnaire format. For example, a need to reduce data entry
time for a survey with many responses may suggest simultaneous
use of more than one operator entering data in separate
computers. This would create two or more data files which would
ultimately require combining in order to analyze the total of all
responses.
DUPLICATNG PROFILE AND DATA FILES
Since, in this situation, you will be merging data files
associated with the same profile file, it will be necessary to
create duplicate profile and data files with root names changed
from the originals so that they can appear in the same directory
for merging. This is accomplished by selecting the SURVEY
DUPLICATE option in the CREATE window of the main menu. If you
have not already selected the directory holding your files, you
will be asked to do so. You will then see a list of the existing
profile files in the directory and should select the one you wish
to duplicate. You will be asked for the new root name for the
duplicated profile and empty data file. With this accomplished,
the program will create the newly named duplicate profile file and
a corresponding empty data file. This will return you to the
main menu.
If the questionnaire duplication is being made to permit
entry of response data simultaneously in separate computers to
speed up processing, one set of matching profile and data files
will have to be transferred to the additional computer. When the
time comes to merge the data files they should be transferred back
to the original computer and directory.
The data files can then be merged by running SURVEY and
selecting FILES COMBINED in the CREATE window of the main menu.
Follow the directions on the screen. When asked, enter one of the
names of the identical profile files. When prompted, follow by
entering, in turn, the names of the two data files to be combined.
After you have verified the files to be combined, the program will
create a file named COMBINED.DTA and a duplicate of your profile
file, named COMBINED.PRO. The new profile file is needed to
permit you to view or print the new combined report.
32
MAKING AN ADDITIONAL QUASI-DEMOGRAPHIC SEPARATION
The file merging capability of the program has another
application in the case where it is desired to make an analysis
by demographic separation, (employees grouped by years of company
service, for example), plus an additional division into separate
groups, (the respondent's gender, for example). Since SURVEY
accommodates only one demographic question, the following
procedure will go a long way toward producing an analysis of
another demographic grouping. It permits the use of a second
quasi-demographic question which must also be in the single
choice, Type A category. It can be placed at any point in the
questionnaire, although it may be helpful to place it last to
simplify a manual sort operation.
When the responses are received in this application, they
should be manually divided into separate groups, based on the
reply to the selected quasi-demographic Type A question, gender
for example. In this case, there are only two such groups.
Before proceeding further, the SURVEY program should be run and the
profile file for the questionnaire created as described earlier.
Then a duplicate profile and empty data file should be created with
a different root name by selecting the CREATE - SURVEY DUPLICATE
option as described in the previous section.
With two differently named but otherwise duplicate sets of
profile and data files created, the responses from each of the
previously separated groups should be entered, each using a
different set of the duplicate profile and data files. A separate
report can then be viewed and printed for each group. Both data
files should be copied to a different directory or disk for
backup. As a second step, the two data files should be merged as
described in the previous section. A combined data file will be
created, named COMBINED.DTA. It will then be possible to print a
report representng the replies of all groups, including a report
for each of the main demographic groups.
The result of the complete procedure is to produce the
following analysis:
■ By gender (in this example), total and by demographic
subdivision
■ By total of each demographic group
■ By total of all responses
33
While the example discussed involved only two quasi-
demographic groups (gender in this case), more than one can be
accommodated in the same manner. This procedure will be simplified
if the question with the minimum number of choices (gender for
example) is used for the quasi-demographic separation, and the
question with the larger number of possible choices (age group for
example) is used for the real demographic separation as Question
Number 1.
DELETING A RESPONSE - In the event of Duplicate entry
It may never happen to you, but a few users have been
dismayed to find they have entered the contents of a response
sheet twice. For such a situation, the program provides the
capability of removing all the entries in a particular response.
This option is accessed in the DATA IN menu window by selecting
DELETE A RESPONSE. You will be asked to enter the response data in
the same manner as if you were adding a new response. Note
however, that "Deleting a Response" will be flashing in the upper
right of the screen. Continue making the entries EXACTLY as they
were originally entered. At the conclusion of this action, the
program will remove the duplication.
INTERPRETING THE CORRELATION REPORT NUMBERS
If your report includes a summary examining the correlation
of question choices in two different questions, you will obtain a
print of number scores and percentages for two conditions; (1), the
cases in which both the Influenced and Inflencing choices were
selected and, (2) the cases in which the Influenced choice was
selected, but not the potentially Influencing choice. The scores
report the total for each occurrance. The percent number is
related to the total of scores (1) and (2).
CHOOSING A SAMPLE SIZE
The total number of individuals in the population being
considered for an opinion survey usually determines whether an
attempt should be made to contact almost all members of this
population, or whether taking a sample is a more practical
procedure. Sampling, however, reduces both accuracy
(reliability) and confidence (raises risk). For a total
population of under one or two hundred, an attempt should be made
to obtain responses from most of the members in order to obtain
trustworthy results. The primary difficulty in getting good
results is obtaining a reasonably random and unbiased sample.
34
Typical objectives for a sampling survey are to obtain an
accuracy of plus or minus 5 percent with 95 percent confidence.
This means that if a truly random survey was made of the same
population 100 times (heaven forbid!), 95 percent of these
surveys would produce results which would not be different from a
survey of the entire population by more than 5 percent.
Some of the references in the following section provide
more detail on the statistics of sampling. As a practical
matter, however, for populations less than one or two hundred,
try to contact as many as possible. For populations in the order
of 500 and a truly random sample, you should contact at least
about 230 to reach the above objectives. For populations of
about 1000, try for about 280. A population of about 10,000
would require about 370 in the sample for the same objectives.
REFERENCES
The following books can provide more information on the
design and application of opinion surveys.
1. Converse, Jean M., Handcrafting the standardized
questionnaire
Sage Publications, Beverly Hills, CA, 1986
2. Blankenship, Albert B., Professional telephone surveys
McGraw-Hill, New York 1977
3. Blankenship, Albert B., Consumer and opinion research
Arno Press, New York 1978
4. Oppenheim, Abraham Naftali, Questionnaire design and attitude
measurement
Basic Books, New York 1966
5. Sonquist, John A., Survey and opinion research
Prentice-Hall, Englewood Cliffs, NJ 1977
6. Stephan, Frederick S., Sampling opinions; an analysis of
survey procedure
Wiley, New York 1963
35