You are on page 1of 6


Pick a relatively general attitude, belief, or behavioral pattern one that could be measured in a
variety of ways. This concept must be at least ordinal: That is, there must be some single
dimension which you can have more or less of.
An example could be money. Money is something everyone has except poor course lecturers
and it is something that we can ask a lot of questions about. Another example could be shopping
something everyone does and something we can ask peoples preferences about. Exercising is
another example. With these concepts, we can ask more or less questions there is a single
dimension that you can have or do more or less off.
So, all of your questions in the survey are focused on measuring this one concept. E.g. 10 questions
about peoples behaviors concerning shopping, or 10 questions about exercising.
All concepts fulfill this criterion, so it is not likely to be a problem for you, but ask if you are not

Select a few (25) independent variables for which it is easy to write simple closedended questions.
Prepare these simple questions and include them in your questionnaire.
Choose one independent variable that you really believe affects your dependent variable; this
variable will be the one you want to use for your hypothesis. For example, feeling guilty over not
exercising may prompt people to exercise (hypothesis: Feeling guilty over not exercising will cause
people to exercise often); or having friends who exercise may cause people to exercise.
Alternatively, getting feeling good after exercising may cause people to exercise more.
The IV may be a basic background variable, or may itself be an attitude or behavior that you can
measure with one or two items.
Pick a few other IVs that you think are background factors you suspect will make a difference in
your dependent variable (age, whether ones friends exercise regularly or not, etc.). These are to give
you the possibility of having additional information if you need it.
These IVs can be nominal, ordinal, or interval (ratio data rare in surveys, but e.g. age form an
exception this is more demographic than actual survey question on the variables though).
Additionally, you will write 12 general openended questions to capture the persons opinion in his
or her own words; we will use this question as a validity check.

Your task is to develop a variety of closedended questions which can be summed to form an index
which measures the DV (chosen subject) with less error than the individual items would have. I.e.,
we ask people 10 questions about their exercise habits rather than just one, to get a more accurate
and well-rounded result.
A. Closedended questions.
There should be 10 closed questions to measure different aspects of the same dependent variable.
You should generally have four to seven response categories for each question, although there are
One of the questions should be a straightforward general question to capture the main idea of what
you are interested in (the DV) (e.g., in the case of exercising: Do you like to exercise?).
The response categories for the questions should be ranked to express more or less of the attitude
you are measuring (for example: ranking questions on a scale from 1-5, where 1 = not a lot and
5 = a lot, and 3 = neutral response option.
Create 9 other questions on different dimensions or themes relevant to the dependent variable
(using the IVs as a guide). In the case of exercising, this could be time (for how long has the
person been exercising?), exercising alone or with friends, types of sports, how long one round of
exercise is, if the person is overweight, willingness to start exercising, etc.
Unless this is impossible, some of these items should be worded positively so that agreeing means a
person is at the high end of your concept (I have been exercising for a long time and I plan to
continue agreeing = High score on Exercise Scale), while others should be worded negatively so
that agreeing means a person is at the low end of your concept (I dont exercise regularly right now
and I dont care agreeing = low score on Exercise Scale).
Our use statements that the respondent can choose to agree or disagree with, e.g.:
1 = Strongly Disagree
2 = Disagree
3 = Undecided
4 = Agree
5 = Strongly Agree
Or a frequency scale, e.g.:
1 never
2 sometimes
3 usually
4 - always
We use the scores to calculate statistical values. For example sample means etc.















needsimmediateattention)Iwillusuallyskipmyexerciseforthatday. 1234






Check the survey: You will end up wanting to add up these items, so that you can give people
one score, which is the sum of their responses, so it has to make sense to do this. In the above
example, the scores for all questions can be added together to get an idea about peoples
relationship with exercise (remember to invert scores for 9 and 10).
B. Openended question. Include in your questionnaire one or two general and straightforward
openended questions that ask people to use their own words to tell you what they think about the
issue that is the subject of the dependent variable.
When evaluating the open-ended responses, try and categorize the answers thematically according
to peoples general attitudes; this will provide useful information for evaluating your closedended

Refine your questions to make them as good as you can, both in their content and in the physical
structure of your questionnaire. All questions should meet formal criteria such as:
A. Unbiased; if biased items are used, they should be balanced.
B. Clear, unambiguous, singlebarreled, grammatical.
C. Closedended categories are exhaustive, mutually exclusive, reasonable in range and precision,
and fit with the stem of the question.
D. Legible, logical physical presentation.

1. Make copies of your teams questionnaire. Each team must collect data from a minimum of 10
people (ideally everyone on the course, but 10 is ok).
2. You may ask the questions orally or let people write their own answers.
3. Put a unique number on each questionnaire, right on the actual paper used for data collection. Use
these same numbers on your datasheet.

1. Enter the data in SPSS in a data summary table. Use rows for respondents and columns for
questions (i.e. variables)
2. Go over the questionnaires and summary table, discussing them and writing notes on them about
problems or inconsistencies which can be seen within one persons answers.
For example, if one person answers positive towards something in one question, negative in
another, you may want to remove this person from the survey or at least keep track of the
information for later use.
If you have the questionnaires, you can write these notes on them, or on a separate sheet of paper,
whichever seems better. Write notes on questionnaires where the pattern of answers seems odd, and
keep track of this information for later use.

4. Inspect Your Data

4.1. Frequency distributions: You can in SPSS make univariate frequency distributions for all
variables (frequency diagrams for each question) e.g. look at the frequency distribution of
different job types.
There are several useful things to inspect the frequencies for.
A. Problems of low variability. If 80% or more of the cases are in one category of a variable, the
variability of that variable is so low that statistical results with that variable are extremely
problematic with small samples. Results from such variables cannot be trusted, and they are often
simply discarded.
B. Lesser problems of low variability, where 60% or more of the cases are in one category, or
where 80% or more of the cases are in two adjacent categories (of a variable with more than three
categories). These may just indicate a skew in the population, but might also indicate a biased
sample or a biased question.
C. Comparisons across the different questions measuring the dependent variable, to see which
items elicit the most favorable responses, and which the least favorable.
D. Information about the distribution of attitudes in the population. Because your samples are
nonrandom, you have to interpret these results very cautiously, but it is still interesting to find out
what people said.
4.2. Bivariate analysis (scatterplots or bar charts two variables along the X and Y axis
respectively) Relations between your various questions measuring the dependent variable should
correlate. If either your general question or your coded openended question is good, this will give
you a quick check on the validity of the other questions.
3. Correlations and reliability analysis among the questions measuring the dependent variable.
Use Pearsons r or Spearman rho as appropriate) e.g. try and correlate job type with number of
times per week going shopping.
Ideally all correlations should be moderately large and positive; negative correlations are definitely
bad unless you are specifically trying to find negatively correlated trends.
Recall that you are trying to make a survey of 10 questions measuring the same thing, but using
different questions so the responses should correlate.
Some typical problems:
A. One or two questions account for all the negative correlations and/or most of the small
positive ones. This probably means those questions are bad (do not measure what you thought
they did) and the others are OK. You discard the bad questions and get a revised index using only
the good ones.

C. Most correlations are moderately or strongly positive, there are no negative correlations and
only a light scattering of weak positive ones: This is good the survey is working!
D. Most correlations in the table are close to zero, and negative ones are scattered across
different questions. This means that none of the items are properly related to any of the others, that
there really is no single concept that your questions measure. Bugger!
E. There are a lot of strong negative correlations for one or two variables. This usually means
that you have forgotten to reverse score the question items, or that subjects read the opposite
meaning into a question than you intended.