SOCI 211 Lecture Notes - Lecture 10: Sociological Inquiry, Official Statistics, Cluster Sampling
3 May 2018
School
Department
Course
Professor
SOCI 211 – Sociological Inquiry – Week 10
Unobtrusive/Non-reactive research
• By definition, does not influence the subject
A pole ith epeietatio, sues, ad ualitatie eseah…
• They all intrude to some degree in the subject of study
o Experimentation: The concerns with experimentation are challenges to internal
and external validity, each of which may be compromised by the experimental
intervention
o Surveys: Similar concerns are posed by survey research:
▪ Do questions cause someone to think about a problem for the first time?
▪ Do close-ended questions, without filters, force respondents to express
opiios the dot atuall hold?
▪ Are responses shaped by concerns about conforming (registering to vote)
or reporting norm violations (drunk driving)?
▪ If an interviewer develops rapport with a respondent, will that rapport
influence the answers provided?
o Qualitative research: It is likely that gang members, those living in public housing
units, and work groups change their behaviour when in the company of a
qualitative researcher
Unobtrusive (or non-reactive) research methods
• B&R (textbook) treat these methods as being non-reactive:
o Analysis of existing statistics
o Content analysis
o Historical and comparative research
• To these can be added various kinds of unobtrusive observations
Kinds of non-reactive research
• Official statistics (government statistics)
o Analyzing data from Stats Canada
▪ Does not mean subjects have not been influenced by data collection, but
research is non-reactive at the step that gets to us
o Distinction between use of statistics with secondary data analysis (though both
are non-reactive)
▪ Secondary data analysis
• Data collected by others
• ‘eseahes o statistial aalsis
• Often done in quantitative research
▪ Use of existing statistics
• Data collected by others
find more resources at oneclass.com
find more resources at oneclass.com
• Ma e soeoe elses statistial aalsis – though descriptive
statistics from a source can often be combined for statistical
analysis
• Used in both quantitative and qualitative research
o Problems:
▪ Official statistics are themselves the result of a social process
• E.g. A crime – Have to go through a whole set of steps before
something appears in police ledgers/police data
o So, incidence of crime is underestimated because there is
slippage in the social process that generates the statistics
▪ Statistics agencies which generate this data may have different practices
• E.g. Stats Canada and the statistics organization in the U.S. may
define unemployment differently
▪ What statistics mean depends upon institutional design
• E.g. Drop-out rates are often seen as bad; but failing to graduate
from high-school means different things in different countries
o Dropping out in Canada has worst labour market
consequences than dropping out in Australia
▪ Different data sources have different strengths and weaknesses
• Cesus gets eeoe, sues dot
• Surveys involve a lot of information
o But there are non-response rates that vary by category of
the population
▪ Often, official statistics are presented as numbers for aggregates
• The difficulty with this is that sometimes this will open
researchers to the ecological fallacy
▪ Validity of the measures may be problematic
• E.g. Assessing quality of the labour market – is it robust or weak?
o Using unemployment rate is an imperfect measure of the
state of the labour market
▪ Some significant proportion of people in a weak
laou aket see that the at get a jo ad
stop lookig = disouaged okes dot sho up
in the unemployment rate)
▪ Reliability will sometimes be a problem
• E.g. Comparing rate of sexual assault over time may be a problem
as laws defining this change over time
• Dot line in a table between time points = change in practices
o When you use statistics, need to look for breaks in series
(change in definition, practices, etc.)
• Content analysis
o The study of recorded human communications
find more resources at oneclass.com
find more resources at oneclass.com
▪ You take written or screened material and analyze the content of that
material
• Ex. Are popular French novels more concerned with love than
Canadian novels?
• Ex 2. Was the popular British music of the 1960s more cynical
than German?
• Ex 3. Looking at politiias speeches – who is more likely to get
elected?
▪
o Steps:
▪ 1. Have to sample
• Determine unit of analysis and/or observation
▪ 2. Have to determine your sampling frame
• What kind of words/communications are you sampling/looking
for?
• From what population?
▪ 3. Use one of the sampling methods we have discussed
• Random, systematic, stratified
• Often use cluster sampling
▪ 4. Gather your data
▪ 5. Code
• Two kinds:
o Manifest coding
▪ Take something readily identifiable and count the
number of times it appears in the text
• Ex. Count kisses/hugs/loe i a oie
▪ More reliable, but less valid
• Reliability – Fairly straightforward to count
words/acts in a book/movie
• Validity – Less apparent that they indicate
what you want them to indicate (ex. words
in different contexts change the meaning)
o Latent coding
▪ Requires making judgments about the content of
the ouiatios that oue aalzig
• Ex. Ask your coders to make judgments
about the overall romantic-ness of a movie
▪ More valid, but less reliable
• Reliability – Different judgments may vary
across coders
• Validity – Judgments have face validity
o Advantages:
▪ Not very expensive
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
Soci 211 sociological inquiry week 10. Unobtrusive/non-reactive research: by definition, does not influence the subject. Unobtrusive (or non-reactive) research methods: b&r (textbook) treat these methods as being non-reactive, analysis of existing statistics, content analysis, historical and comparative research, to these can be added various kinds of unobtrusive observations. Stats canada and the statistics organization in the u. s. may define unemployment differently: what statistics mean depends upon institutional design, e. g. Are popular french novels more concerned with love than. Was the popular british music of the 1960s more cynical than german: ex 3. Looking at politi(cid:272)ia(cid:374)s(cid:859) speeches who is more likely to get elected: steps, 1. Have to sample: determine unit of analysis and/or observation, 2. Have to determine your sampling frame: what kind of words/communications are you sampling/looking for, from what population, 3. Use one of the sampling methods we have discussed: random, systematic, stratified, often use cluster sampling, 4.
