Psychology Lab – Methods in Behavioural Research: Cozby.
Survey Research. CHAPTER 7.
Goal is to ask people about themselves. Attitudes/beliefs.
Provides us with methodology to ask of them.
Surveys can be used for public policy services.
Study relationships btwn Variables such as attitudes/emotional states/self-reports
statements/ come into play.
Survey research – complement to experimental research and multiple methods are
needed to understand behaviours.
Response set – take into interest the bias’ of people as they answer to see the
tendency to respond to questions from one perspective. Social desirability is the
Conduct questions to Ask
Defining Research objectives: through attitude and belief changes/facts and
Question wording: a) unfamiliar technical terms b) vague terminology c)
ungrammatical sentence structure d) phrasing overloads working memory e)
embedded question with misleading information.
Avoid – double barreled questions (asking two things at once), loaded questions,
negative wording, “yea saying/ney saying”- may be agreeing with you are saying.
Responses to Questions
Closed Versus/Open ended questions:
oCV = with multiple choice and OE = is open ended.
Response alternatives – Strongly agree to strongly disagree.
Rating scales to Graphic Rating Scales. (end points label extremes- graphic has
checks measured on a line).
Semantics Differential Scale (is any concept respondents rate from persons, objects,
beh. Etc using bipolar adjectives – on a 7point scale).
Nonverbal Scales for Children. (happy faces)
Labeling response alternatives.(when asking about frequency of beh. Orrr to define
meanings of each alternatives).
Formatting the questionnaire. (no errors/space/neatly/ use scales consistently ]
[ sequence of questions – themes]).
Refining Questions (think aloud procedure – see how they interpret question)
Two ways (WRITTEN//INTERVIEW FORMAT)
oQuestionnaries (less costly than int’s/anonymous/--- bad must know to
read/might be boring/motivation.
oPersonal Admin to grps/ind. (advantage is usually the audience is more
oMail surveys (inexpens way/low response rate).
oInternet Surveys (open + closed ended questions –newsgroups/email
invitation/chat rooms/bulletin boards).
oPagers/cellphones/or with emotions.
oMore likely to agree to answer for real person/response rates are
hugher/rapport is built btwn the two/answ q’s that need verifying/interviewer
bias may be an issue.
oFocus groups (6-10 ppl for abt 2-3hrs). Usually incentives to
participate/requires time and cost traveling).
Survey Designs studying Changes over time
For comparison purposes. Changes in attitudes/aspirations of beh/composition.
Another way to study over time is (panel study – same ppl at two or more points in
time. They have waves – representing amount of surveys).
Sampling frm a Population
Pop is composed of indiv of interest to the researcher. Sampling involves getting ppl
from this population of interest to survey.
Sampling allows us to estimate the characteristics of the pop as a whole. Stats
theory allows us to infer what the pop is like based on data obtained from the
sample. Logic of this is called the statistical significance.
Saying that this is accurate within 3% points. A score of 61% would mean in btwn
58-64%. 95% of the population lies within this interval.
Sampling errors occur so this is a margin of error. May deviate due to measurement
Larger sizes will reduce the size of the confidence interval.
Sample size is the single greatest factor of intervals.
Sizes are calculated using conservative assumption about nature of pops.
Two techniques for sampling individuals from a population – probability sampling
and nonprobability sampling.
oEach person of the pop has a specifiable prob of being chosen. P.S. is
important for precise statem’s about a specific pop on the basis of results on
ur survey. With simple random sampling every member of pop has = prob
of selection. Stratified Random s is divided into subgrps and are used to
select sample memb from stratum. Surveys can do this based on
age/gender/education for sexual attitudes. Small grps in pops will be
oversampled to make sure they get an accurate response. Cluster sampling
occurs when u form a group (cluster) to sample from a list of individuals.
Once clusters are chosen, all individuals in each cluster are included in the
oDon’t know about the prob of any member of the pop being selected. Quite
common and useful alothough not as sophisticated as prob. Sampling.