Midterm 1 Flashcards
Advantages of open questions
respondents can answer in their own terms
allows unusual answers
allows tapping into participant’s knowledge
good for exploring new areas
disadvantages of open questions
time consuming to record/code
length may put respondents off
inaccuracies in transcription of spoken answers
coding, and the types of coding
deriving themes/categories of behaviour. researcher usually assigns number to code
allows information to be coded quantitatively
–> pre coding and post coding
post coding
going back to info to look for incidences of theme or category, may be unreliable because of inconsistencies in judgement from different coders
pre coding
when researcher designs coding grame before administering the survey
3 basic principles of coding
categories are mutually exclusive
categories are exhaustive (including “other”)
clear rules regarding how codes are applied (ensuring consistency)
advantages of closed questions
easy to process easy to compare set of answers help clarify the meaning of the question quick and easy to complete reduces risk of bias from recorder
disadvantages of closed questions
answers lack authenticity/spontaneity
care needs to be taken to prevent overlap in categories
difficult to make answers exhaustive
irritates respondents when answer categories aren’t relatable
reduces conversation/rapport in interviews
types of questions
- personal factual questions (age, occupation, how often do you go to the movies? etc. often have to rely on memory to answer)
- factual questions about entity or event (good when info isn’t available elsewhere, leads to problems because people aren’t careful/systematic observers)
- questions about beliefs (should Canada maintain military presence abroad?)
- questions about attitudes (common in structured interviews/questionaires, Likert scale is common)
- questions about knowledge (who was Canada’s first prime minister?)
general rules for designing questions
- keep research question in mind (reduces risk of asking irrelevant questions)
- being specific (what exactly do you need to know?
- recognize ambiguity (how would you answer it?)
what does it mean to avoid ambiguity?
- avoid “often” and “regularly” as measures of frequency
- clarify words that mean different things to different people (ex: dinner)
what does it mean to avoid long questions?
-long questions may be nice for questions asking about behaviour, the longer they take to answer, the more it may facilitate memory recall
what does it mean to avoid double-barrelled questions?
ex: how frequently do you cook and clean? respondent may cook but never clean
what does it mean to avoid general questions?
.
what does it mean to avoid leading questions?
.
what does it mean to avoid questions that ask several questions
ex: who did you vote for in the 2011 federal election? should be did you vote in the 2011 federal election? if yes, which party did you vote for?
what does it mean to avoid negatives?
confusion can lead to innaccurate answers
what does it mean to minimize technical values?
.
what does it mean to ensure respondents have requisite knowledge?
-if respondents do not know about topic, answers won’t be meaningful
what does it mean to ensure symmetry between question and answer set?
ex: do you believe in God? strongly agree, agree, disagree, strongly disagree
what does it mean to ensure answer set is balanced?
equal number of positive responses to negative responses
what does it mean to not overstretch people’s memories?
ex: how many times do you drink in a year? vs how many times do you drink in a week?
what does it mean to provide “don’t know” options?
- avoids forcing expression of views that aren’t held
- allows out for those too lazy to do thinking (lower education, later questions in survey are more likely to utilize ‘don’t know’ option
what does it mean to consider question order?
- all respondents should receive questions in the same order?
- researchers should be sensitive to possible effects of order
tips on question order
- general questions before specific questions (if specific is before general, aspects of specific answer may be omitted from answer for general because it has already been addressed)
- opinions/attitudes before behaviour/knowledge (ex: spouse reports doing 20% of housework, influences question ‘is housework shared equally?’)
- early questions should be related to announced topic
- important/meaningful questions early to stimulate interest
- questions that cause embarrassment/anxiety go at the end but not the very end or else the interview will leave a negative impression
- personal questions that are apparently irrelevant to research question go at the end
- long questions should be grouped by topic
- if respondent answers something that will be asked later in the interview, still ask the question because answer might’ve changed
vertical vs horizontal answers
- vertical reduces likelihood of confusion
- vertical clearly distinguishes answers from questions
vignette questions
-a form of closed question used to examine ethical standards and beliefs by presenting scenario(s) and asking how participants would respond
advantages of vignette questions
- anchors responses in realistic scenario, reduces unreflective reply
- because vignette is about other people, allows for distance between question and respondent, leading to more candid replies
disadvantages of vignette questions
- impossible to establish assumptions made by respondent about the scenario
- difficult to establish how far from respondent’s normative views (ie what people say they’ll do and what they do are different)
pilot studies
- pilot study cannot use members of sample in true study
- if study will use closed questions, open questions can generate closed answers
- can help develop interviewer’s confidence
- helps identify unuseful question, highlights needed modifications
- flags questions that make respondents bored/uncomfortable
- identifies questions most often skipped
- determines adequacy of instructions
- offers opportunity to evaluate overall flow
using existing questions
- already piloted, tested for reliability/validity
- allows for comparisons to other research
- provide insight on best way to approach research
the main steps in quantitative research
theory, hypothesis, research design, devising measures of concepts, select research site, select research subjects/respondents, administer research instruments/collect data, process data, analyze data, findings/conclusions, write up findings/conclusions
concept
ideas or mental representations of things
- building blocks of theory
- represents points around which social research is conducted
- categories for organization of ideas/observation
- concept can be interdependent or dependent variable, descriptive or comparative
independent variable vs dependent variable
something to be explained vs possible explanation
concept can be descriptive or comparative
changes in amount of social mobility in Canada over time vs variations among comparable nations in levels of social mobility
why measure concepts?
- allows for delineation of fine differences between people in terms of characteristic in question (it’s harder to recognize fine distinctions than extreme differences)
- Provides consistent device for gauging distinctions (measure’s results shouldn’t be affected by time/person administering the measure)
- Provides basis for estimates of the nature/strength of relationship between concepts
Indicators
stand for or represent concept, necessary to measure concepts (can be indirect, for example absenteeism as an indicator for low job satisfaction)
two types of definitions of concepts in quantitative research
- nominal; describes in words like dictionary (crime is any violation of the Criminal Code of Canada)
- operational; spells out operations that will be performed to measure concept (to measure crime, this researcher will use statistics provided by police force)
ways to devise indicators
through questions part of the interview/questionnaire (respondents attitudes, personal experiences, behaviours, etc)
developing criteria for classifying observed behaviour (pupil behaviour in classroom)
through use of official statistics (stats canada)
developing classification schemes to analyze written data (analysis of how newspapers characterize sex workers)
using multiple-item measures in survey research
single indicator may misclassify some individuals if wording leads to misunderstanding of meaning
single indicator may not capture all meaning in underlying concept
multiple indicators allow for finer distinctions and sophisticated data analysis
reliability
concerned with consistency of measures by looking at stability over time, internal reliability, and inter-observer consistency
stability over time
- whether results fluctuate as time progresses, assuming that thing being measured isn’t changing
- most thermometres have this reliability
- test using Test-retest method
internal reliability
- aka consistency
- multiple measures administered in one sitting should be consistent
- cronbach’s alpha coefficient
- split half method
cronbach’s alpha coefficient
commonly used test in which 1 is perfect internal reliability and 0 is no internal reliability, and .8 is typically considered minimum acceptable level
split half method
indicators divided into two halves, respondent’s scores should correlate, in which 1 is perfect internal reliability and 0 is no internal reliability
inter-observer consistency
- judgements between several researchers in activity involving subjective judgement
- ex: classifying and categorizing open answers
measurement validity
whether indicator accurately/properly gauges concept
-face, concurrent, construct, convergent
face validity
measure appears to reflect concept, essentially intuitive process
concurrent validity
examining when criterion differs from case to case
-ex: when measuring absenteeism as indicator for low job satisfaction, lack of absenteeism should be seen in those with high job satisfaction
construct validity
whether concepts relate to each other in a way consistent to what their theories would predict
ex: routine jobs should have lower job satisfaction than jobs with varied activities. if the routine jobs are found to have equal job satisfaction as complex jobs, it lacks construct validity. Either measure was invalid, deduction was misguided, or theory needs revision
convergent validity
- validity should be gauged through comparison to other measures of same concept developed through different methods
- ex problem with convergent approach: measuring crime though police reports or victimization surveys
which validities are more important?
face and internal are usually the only ones tested
if measure is not reliable ….
it cannot be valid
goal of quantitative researchers
to understand social order by making sense of phenomena and evaluate theories and interpretations
establishing causality
describing why, not just how
good quantitative research inspires confidence in researcher’s causual inferences
generalization of findings
sample must be as representative as possible in order to be confident results are not unique to the sample
probability sampling largely eliminates bias through random sampling
critiques of quantitative research
- fails to distinguish people from ‘world of nature’ (some claim science is only applicable to entities/processes that lack self-reflection)
- measurement process produces false sense of precision/accuracy
- reliance on procedure creates disjuncture between research and everyday life (relates of external validity, difference between what people do/say they’ll do)
- analysis of relationship between variables promotes view of social life as remote from everyday experiences
- explanations for findings may not be empathetic (ex: poor inner city areas see more unwed mothers with children because women are marrying later in life instead of marriage losing its popularity)
- assumption of objectivist ontology (assumes reality exists)
document
can be ‘read’, was not produced for social research
- analysing documents is unobtrusive and non-reactive
- removes threat of validity
4 criteria for assessing quality of documents
- Authenticity (genuine and unquestionable origin?)
- Credibility (free from error and distortion)
- Representativeness (typical of time/place? extent of uniqueness?)
- Meaning (clear and comprehensive?)
diaries, letters, autobiographies
- often used by historians, not social researchers
- written by purported author?
- people who are aware of an audience may not reveal everything on paper (what people don’t write can be of significance)
- who was able to record/write/read?
- what has been damaged/destroyed?
3 types of household photograph
idealization (formal portrait)
natural portrayal (informal snapshot)
demystifcation (subject in atypical, often embarrassing situation)
government documents
authentic, have meaning
perhaps not credible
advantages of secondary analysis
- saves costs and time
- high quality data (sampling procedures are usually rigorous, often national scope, generated by highly experienced researchers)
- opportunities for longitudinal analysis
- subgroup analysis (when samples are large)
- opportunity for cross-cultural analysis
- more time for data analysis
- reanalysis can offer new interpretations
- wider obligations of social researcher (social research is chronically underanalyzed)
disadvantages of secondary analysis
- lack of familiarity with data
- complexity of data
- ecological fallacy (occurs when you make conclusions about individual based on analyses of group data)
- no control over data quality
- absence of key variables
advantages and disadvantages of official statistics
- often based on whole populations rather than samples
- problem of reactivity is less pronounced
- possible to chart trends over time
- but records only those processed by stats collectors (consider crime/suicide rates)
unobtrusive method
removes observer from behaviour under study
-studying physical traces (like grafitti, paper trail of finance), archive materials (data form governments and ngos), and simple observation (observer has no control over behaviour)
types of variables
nominal, ordinal, interval/ration
nominal variables
- aka categorical, composed of categories with no relationship except that they are different
- order of categories is arbitrary
ordinal variables
- categories that can be ranked
- can be described as
- likert scale is common
- difference between categories is not necessarily equal
- no unit to measure
interval/ration variables
- can be measured by unit
- difference between categories is equal
- can have 0 value
- can be ranked
frequency tables
provides number and percent of subjects belonging to each category of variable
measures of central tendency
mean, median, mode; provides typical score in one number
mode
value that occurs most frequently, applicable to all types of variables, especially nominal data
median
mid point of scores, if there is an even number of scores the median is the mean of the middle 2 values. applicable to interval/ration and ordinal variables
mean
average, vulnerable to outliers
range
difference between the highest and lowest value, vulnerable to outliers
standard deviation
variation around the mean, vulnerable to outliers
work out the general mean, subtract the mean from every value, square every value, then find the mean of those values
bivariate analysis
examines relationship between 2 variables, esp through use of contingency tables
pearson’s r
statistic used to examine relationship between 2 interval/ratio variables
the relationship must be broadly linear
statistical significance
indication of risk of genralizing sample statistic to population
set up null hypothesis, establish acceptable level of statistical significance, determine statistical significance, decide whether or not to reject the null
two types of error
type I - true null was rejected
type II - false null wasn’t rejected
chi-sqaure test
applied to contingency tables
- measure of likelihood that relationship between variables in sample will also be found in population
- large chi-square to reject null hypothesis, larger n makes this easier
spurious relationship
when relationship appears to exist but isn’t real
intervening variable
suggests relationship between 2 variables is not direct
prominent sources of error
- poorly worded questions
- interviewer error in asking question
- interviewee misunderstanding question
- interviewer lapses in memory
- interviewer error in recording answer
- mistakes entering data into computer
- bias caused by innate characteristics of interview(er/ee)
- intra-interviewer variability
- inter-interviewer variability
advantages of phone surveys
- cheaper and quicker
- easier to supervise
- confidentiality is not as much of an obstacle
- reduces bias because interviewer and interviewee can’t see each other
disadvantages of phone surveys
- can only reach those with listed phone number
- hard to communicate with hearing impaired
- cannot be sustained for long times (
advantages and disadvantages of computer assisted interviewing
- good for filter questions
- good for randomizing order
- loss of sponteneity
- lack of rapport/commitment/motivation
research driven diaries
record of feelings/actions shortly after they occue
advantages and disadvantages of research driven diaries
- accurate data
- expensive
- attrition
- details may not be recorded quickly enough
advantages of questionnaires
- cheaper
- quick
- easier
- absence of interviewer effects
disadvantages of questionnaires
- cannot explain question
- greater risk of missing data
- cannot probe
- difficult to ask a lot of questions
- questionnaire can be read as a whole
- cannot verify who took the survey
problems with respondents
- acquiescence (agreeing just to please the interviewer, ex: bodychecking should be banned from hockey)
- social desirability
- laziness/boredom
sampling
the selection of individuals/units of analysis
sampling frame
list of elements from which sample is chosen
3 sources of sampling bias
- not using random sample
- sampling frame is inadequate
- non responses
types of sampling probability
simple random sample, systematic random sampling, stratified random sampling, multi-stage cluster sampling
types of non-probability sampling
convenience sample, snowball sampling, quota sampling, structured observation and sampling
reducing non response
-keep calling back, be optimistic, reassure people that you aren’t a salesperson, dress to appeal to wide spectrum of people during in-person interviewing, make sure you are flexible with time
3 Questions to ask during bivariate anlaysis
- does the association exist?
- how strong is the association?
- in what direction does the association exist?
calculating association with bivariate table
“percentage down, compare across”
- an association exists if column percentages change
- the greater the change, the stronger the relationship
- to measure maximum difference, find biggest difference in column percentage for any row of the table
p
probability that results are not due to chance is 95%
null hypothesis
there is no relationship between 2 variables
