Transcript Lecture slides since Exam 3

SOCY3700
Selected Overheads for
the Final Exam
Prof. Backman
Spring 2008
Stages in Field
Research
• Choosing the site
– Start where you are
• Getting in
– Being accepted
– Anonymity
• Getting on
– With self, “the folk”, conscience and
colleagues
• Gathering data
– Logging
– Interviews
• Focusing
• Analysis
• Write up
• (Adapted from Lofland and Lofland)
Street Corner Society:
The Social Structure of
an Italian Slum
William Foote Whyte,
1943 (third edition, 1981)
Whyte Bio
• Educated middle class upbringing
• Loved to write
• Attended Swarthmore in suburban
Philadelphia
• Engaged in some reform activities
in college, but engaged even more
in writing
• Wrote a novel, decided it was lousy
because he didn’t have enough to
say
• Got a Junior Fellowship at Harvard
– three years just to hang around
and do whatever research took his
fancy (sort of)
The Research Problem
• Whyte came to Harvard knowing
mainly that he wanted to study slums
and somehow improve the world
• Social scientific literature was just
beginning to appear. He read lots of it
• Other folks at Harvard had done
similar work and were developing
some theoretical ideas about group
process
– One would not think one would go to a
slum to study group process, but in the
end that was a big part of what Whyte did
• Many of the ideas Whyte when he
started his work came to naught
– “We set out on the frontiers of our
personal knowledge and began exploring
beyond those frontiers” (Whyte 1984:63)
“Cornerville”
• In the usual fashion, Whyte gave
his city and neighborhood a
psuedonym. Cornerville refers to
the slum, now known to be
Boston’s North End. He called
Boston “Eastern City.”
• At the time (around 1937)
Cornerville was suffering the effects
of The Great Depression
• Predominately Italian in a city
whose big politicians were mostly
Irish
• Many residents spoke only Italian
Getting In
• Wandered around Boston,
settled on Cornerville because
it “looked like” his vision of a
slum
• Could observe from the
outside, but wanted to observe
from the inside
• After various failed schemes,
introduced to Doc by the social
worker in charge of girls’
programs at the local
settlement house
• Moved into the neighborhood
Doc
• Doc (a psuedonym for Ernest
Pecci) is probably the most
famous informant in sociology
– A pretty good sociologist himself
for someone who never had a
sociology course
• Late 20s, mostly unemployed
guy from the neighborhood
• Informal leader of a group of
similarly underemployed age
mates
• Interested in making things
better
Doc and Bill
• Doc’s famous response to Whyte’s
first rambling description of what
Whyte was trying to do in
Cornerville:
“Do you want to see the high life
or the low life?”
• Doc served as Whyte’s sponsor,
guide, and “member validator”
– Having a sponsor can be a problem in
settings with a great deal of conflict, as
you may be seen as being on your
sponsor’s side
– “Member validator”: insider who
reviews the sociologist’s analysis from
an insider’s point of view
Getting On
• Whyte moved into Cornerville,
taking a room with a family
• Whyte tried to learn Italian
– Though never got proficient, he
felt his efforts gave him a great
deal of credibility, especially with
the older generation
• Joined various clubs,
becoming secretary of at least
one
• Hung out with Doc’s gang
• Returned regularly to Harvard
for baths and brainstorming
with other social scientists
Going Native
• When you start to act like and
especially to think like the people
you are studying, you have gone
native
– Quite common occurrence
– It is difficult to completely go native
• Whyte’s efforts to swear like the
other guys weren’t successful,
partly because they wanted him to
be himself
– Can get you in trouble
• Whyte voted illegally
• Whyte almost inadvertently got
engaged because he didn’t
understand as much of native
practice as he thought
– The natives aren’t always grateful
Street Corner Society:
Sources
Whyte, William F. [1943] 1981.
Street Corner Society. 3rd ed.
Chicago, IL: University of
Chicago Press.
Whyte, William F. 1984. Learning
From the Field: A Guide from
Experience. Newbury Park,
CA: Sage.
Whyte, William F. nd. Various
personal and classroom
communications.
Bernard on
Unstructured Interviews
• H. Russell Bernard – cultural
anthropologist from U of Florida,
author of a research methods text I
have used in advanced research
methods courses
– As surveys are to sociologists, so
unstructured (and semi-structured)
interviews are to cultural
anthropologists
– As a researcher, journal editor, and
methods text author, Bernard has
been given credit for strengthening the
rigor of anthropological research
Source: Bernard, H. Russell. 1995. Research
Methods in Anthropology: Qualitative and
Quantitative Approaches. 2nd ed. Walnut Creek,
CA: AltaMira. Mostly Chapter 10, pp. 208-36.
Bernard on Unstructured
Interviews (2):
Continuum of Interview
Situations
Since the researcher is an outsider,
the locals will generally be aware
that any contact is likely to involve
information gathering
• Continuum of situations based on
how much the interviewer
controls the situation
1. Informal interview – more or less
normal conversation
-
Typical early in research
Useful for rapport
Useful later for finding topics that might
have been overlooked
Bernard on Unstructured
Interviews (3):
Continuum of Interview
Situations (2)
2. Unstructured interview – not just
normal conversation, but with
minimal control over the responses
of the interviewee
3. Semi-structured – like unstructured
but with an interview guide
-
-
Interview guide: written list of topics,
probes, etc. intended to be covered in
the interview
More formal than unstructured
4. Structured – questions (and often
answer choices) established ahead
of time by the interviewer
- For example, standard survey interviews,
self-administered questionnaires
Bernard on Unstructured
Interviews (4):
Starting the Interview
• Assure anonymity
• Explain their importance to
your understanding
• Ask for permission to record
the interview and to take notes
– The value of the interview much
lower if you can’t record or take
notes
– Even with recorder it helps to
take occasional notes
Bernard on Unstructured
Interviews (5):
Let the Informant Lead
Rule # 1: get an informant on
the topic and get out of the
way
– You pick the topic, interviewee
provides the content
– In general, it is the interviewee’s
ideas you are interested in, not
yours
• This rule is not always slavishly
followed
– Interviewee may stray off topic
– You may have ideas you want
responded to
Bernard on Unstructured
Interviews (6):
Probes
• Use probes to guide interview
• Probe (Bernard definition):
stimulating an informant to give
more information without injecting
yourself so much into the
interaction that you get only a
reflection of yourself in the data
– There are many types of probes
– Our textbook definition: a neutral
request to clarify an ambiguous
answer, to complete an incomplete
answer, or to obtain a relevant
response (p. 192 in Neuman 2007)
Bernard on Unstructured
Interviews (6):
Types of Probes 1
• Silent probe – don’t say
anything when the interviewee
stops
– Difficult to do appropriately
– Culturally sensitive since
different cultures have different
rules about silence
• Echo probe – repeat the last
thing the interviewee said
– Signals that you are interested in
what was said without saying
why or suggesting what to say
Bernard on Unstructured
Interviews (7):
Types of Probes 2
• Uh-huh (neutral) probe –
make regular affirmative
noises, as one often does in
normal conversation to indicate
you are still listening and are
interested
– Keeps the interviewee talking
Rule #2: In general, more
talking by the respondent is
better
– Hence, longer responses are
better
Bernard on Unstructured
Interviews (8):
Types of Probes 3
• The long question probe –
instead of keeping a question
short and to the point, asking a
long roundabout question
– You’re modeling the kind of long
answer you want to get back
– The trick is not to guide the
answer as you ask the question
Bernard on Unstructured
Interviews (9):
Types of Probes 4
• Probe by leading – ask a
leading question as a way of
focusing provoking the
interviewee
– Usually we try not to lead, but
sometimes respondents seem to
be avoiding a topic or conclusion
– Can be used to ask about more
specific incidents or about what
happens when things don’t work
out as expected
– Often based on earlier interviews
Bernard on Unstructured
Interviews (10):
Types of Probes 5
• Phased assertion (baiting)
probe – you take some
information that may or may
not be true and ask questions
as if it were true
– For example, “I guess Hilary and
Barak are friends again. I
wonder why.”
– This is a favorite ploy of gossipmongers
Bernard on Unstructured
Interviews (11):
Verbal Respondents;
Equipment
• Verbal respondents – don’t
be afraid to interrupt a long
winded respondent who is
wandering away from your
topic. Try to be graceful about
it
• Equipment – always make
sure that your tape recorder is
ready before the interview
(fresh tapes and batteries)
Bernard on Unstructured
Interviews (12):
Uses of Unstructured
Interviews
• A primary source of raw data
• Preparation for semi-structured
interviews
• To get info from people unlikely
to give more formal interviews
• Developing rapport
• Studying sensitive topics
– E.g., hot political topics,
sexuality, racial prejudice
– Conflict: you can get wide range
of information from multiple
interviewees
Bivariate
Relationships with
Integer-level Variables
Preliminaries to multiple
regression
Steps in Analysis of
Bivariate Relationships
Between
Integer-level Variables
• Look at scatterplot
– Dependent variable as the Y
(vertical) axis
– Independent variable as the X
(horizontal) axis
• Make best-fit line
– Since it is a line, we call it linear
regression
– Since we have only one
independent variable, we call it
simple linear regression
• Calculate slope (b)
• Calculate goodness of fit (r)
Interpretation of Simple
Regression Results
Equation:
Dependent
= intercept +
coefficient * independent
+ error
• Coefficient (aka b, beta, or
regression coefficient) tells
how many units of the
dependent variable go with the
increase of one unit on the
independent variable
– Mathematically, the slope
Interpretation of Simple
Regression Results (2)
• Correlation coefficient (aka r,
Pearson’s r) – a measure of how
well the line fits the data, usually
interpreted as how strong the
relationship is
– Measures the “goodness of fit”
• The higher the absolute value of r, the
better the fit
– Ranges between -1 and 1
• Positive coefficient means there is a
positive relationship between the two
variables (high on the independent goes
with high on the dependent)
• Negative coefficient means there is a
negative relationship between the two
variables (high on the independent goes
with low on the dependent)
Interpretation of Simple
Regression Results (3)
• Intercept – how many units of
the dependent variable you
would be expected to have
with 0 units of the independent
– Mathematically, it is where the
line crosses the vertical axis
• Error – the difference between
what was actually measured
for the dependent variable for a
particular case and the
measurement predicted by the
equation for the line
Interpretation of Simple
Regression Results (4)
• Statistical significance –
tests how sure we are that the
regression coefficient is not
zero OR that the correlation
coefficient is not zero
– Conventionally we use the 95
percent confidence level
– At the 95 percent confidence
level, the probability of a false
positive is less than 5 percent,
usually written as p<.05
Interpretation of Simple
Regression Results (5)
Example
Dependent variable: violent crimes per
100,000 population
Independent variable: percent of population
15 and up who are currently divorced
Correlation coefficient = 0.24
There is a positive relationship
Regression coefficient = 38.6
For every additional 1 percent to the
percent divorced of the population 15+
there is an increase in the violent crime
rate of 39
Intercept = 160
If no one in the population were divorced,
there would be 160 violent crimes per
100,000
The relationship is significant at the p<.048
level
Multiple Regression
• Multiple regression is multiple
because it allows the use of
more than one independent
variable
– This is nice since so much of
social life has multiple causes
• Multiple regression is probably
the most important statistical
tool in use in sociology today
• There are many similarities
between simple regression and
multiple regression
Multiple Regression (2):
Similarities with Simple
Regression
• The key mathematical operation is
fitting a line to the data points
– The method is the same: choose the
line that minimizes the squared
distances between the points and the
line
• Called the method of least squares;
the line is sometimes called the
least squares line. Sometimes it is
called the ordinary least squares
(OLS) line
• There is a statistic for the overall fit
of the line to the data points
• Each independent variable gets its
own regression coefficient
Multiple Regression (3):
Differences from Simple
Regression
• Scatterplots are in hyperspace
– That is, for each variable, including the
dependent, there is another dimension
in the graph
• They’re really hard to draw!
• The goodness of fit statistic doesn’t
tell you the direction of the
relationships
– We use R (not r) as its symbol
– Actually, we usually use R2
– R2 tells us the proportion of variation in
the dependent variable that is
accounted for by the independent
variables
Multiple Regression (4):
Interpretation of
Regression Coefficients
• New term: ceteris paribus – all
other things being equal
• A regression coefficient tells us
how much change in the
dependent variable is
associated with a change of
one unit in the coefficient’s
independent variable, ceteris
paribus
Multiple Regression (5):
The Regression Equation
• Multiple regression is based on
the matrix equation
Y = XB + e
where Y is the dependent
variable, X is the matrix of
dependent variables, B is a
vector of regression
coefficients (and the intercept),
and e is the error
Multiple Regression (6):
Varieties of Multiple
Regression
• Ordinary regression makes certain
assumptions about the relations
between the independent variables
and about the errors
– These assumptions are not always
met
• Ordinary regression is limited to
only one dependent variable
• There are a large number of
modifications to ordinary regression
that overcome some of its
limitations and to loosen the
assumptions
Multiple Regression (7):
The General Linear Model
• The collection of modifications and
extensions to ordinary regression is
called the general linear model
– The GLM is based on the equation
given earlier
– It brings together a wide range of
statistical methods, some of which had
been invented independently
• The GLM is a conceptual and
methodological breakthrough
paralleled in its importance for
quantitative social science only by
the discovery of sampling theory