-
theory
explanation of observed regularities or patterns
-
what are theories composed of?
- definitions
- descriptions
- relational statements
-
relational statements
connect two or more variables so that if you know the value of one variable you can convey information about the other variable
-
deterministic relational statements
two variables that always go together in a particular way
-
probablistic relational statements
two variables go together with some degree of regularity but the relationship isn't inevitable
-
theories of the middle range
explanations of specific social phenomena (ex. job satisfaction, criminal behaviour, suicide, etc.)
-
grand theories
- broad sweeping historical explanations of societal change
- general and abstract
- (ex. feminism, structural-functionalism, etc).
-
concept
- general/abstract idea
- category that serves to organize observations and ideas about some aspects of the social world
-
deductive approach
- theory -> observations/findings
- researcher comes up with a theory to explain a certain phenomenon, then deduces hypothesis
- if the findings don't support the hypothesis, the theory needs to be revised or rejected
-
inductive approach
- observations/findings -> theory
- a theory is the outcome of research, not the starting point
- you are constructing a theory instead of testing a theory
-
epistemological assumptions
- notions of what can be known and how knowledge can be acquired
- what is considered "acceptable knowledge"?
-
positivism
- an epistemological position
- affirms the importance of following the natural sciences
- values empiricism (only phenomena confirmed by the senses can be accepted as knowledge, subject to empirical testing)
- "value-free" science - scientists from any place or situation if given the same data must be able to come to the same conclusions
-
deduction and induction in positivism?
- deduction: key purpose of theory is to generate hypotheses that can be tested, allow assessments of explanations of observed laws and principles
- induction: can also arrive at knowledge by gathering facts (systematic collection) which leads to a generalization of laws
-
where do positivists thing scientific statements and normative statements belong?
- scientific statements (describe how and why certain social phenomena operate the way they do) belong in science
- normative statements (outline whether certain acts or social conditions are morally acceptable) belong in philosophy or religion
-
interpretivism
- sees the role of social scientists to group the subjective meanings of peoples actions
- people use common-sense constructs to interpret their lives, and these thoughts motivate their behaviour
- interpretivists want to access the 'common-sense thinking' of people to interpret peoples actions and social world from the point of view of the actors
- alternative to the social science usually done by positivists
-
ontological considerations
- branch of philosophy concerned with the nature of reality
- ex. "what kind of things have existence?"
-
If you answer 'yes' to this question what view do you hold: "Do social phenomena have an objective reality, independent of our perceptions?"
- objectivist
- think there is no such thing as social reality
- relation to research: likely to emphasize on formal properties of organizations
-
If you answer 'yes' to this question what view do you hold?: "Is what passes for reality merely a set of social constructions?"
- constructionist
- no objective social reality against which our conceptions and views of the world may be tested
- relation to research: likely to focus on active involvement of people in reality construction
-
values
standard by which we assess each other
-
quantitative research
uses numbers, statistics in collection and analysis of data
-
qualitative research
relies on words, non-numerical symbols
-
influences on social research
- theory
- practical considerations
- epistemology
- ontology
- politics
- values
-
research methods
the logic and techniques of collecting and analyzing data
-
research design
broad structure that guides the collection and analysis of data
-
nomothetic
explanation that applies to humanity in general, not just the people in the study
-
criteria for being nomothetic research
- 1. correlation - proposed cause and effect must vary together
- 2. time order - the proposed cause must happen earlier in time than the proposed effect
- 3. non-spuriousness - alt. explanations for the correlation observed must be ruled out
-
idiographic explanations
- doesn't necessarily apply to others, but helps to explain why the actors of interest behave the way they do
- involve a detailed 'story' or description of people studied based on empathetic understanding
-
3 most prominent criteria for evaluating social research
- reliability
- replicability
- validity
-
types of validity for evaluating social research
- measurement/construct validity
- internal validity
- external validity
-
naturalism
- style of research that seeks to minimize the use of artificial methods of data collection
- social world should be disturbed as little as possible
-
deterministic hypothesis
- true or false
- ex. heat water to 100C and it will boil
-
probablistic hypothesis
- a prediction statement?
- ex. children from poorer homes are less likely to do well in school
-
measurement reliability
- has to do with the stability/consistency of measurement
- an empirical issue
- do you get the same result if you measure again?
-
inter-researcher reliability
- consistency across researchers
- ex. clearly asked questions with no room for interviewers to modify questions
-
inter-observer reliability
- consistency across observers
- clear rules for recording what you are observing
-
measurement validity
- are you really targeting what you want to be measuring?
- the adequacy of the measurement -> conceptual issue
- ex. does the number of rooms in a house measure wealth? not really
-
element or unit
single case in the population
-
population
all cases about which you are seeking knowledge, or all the cases to which your conclusions are meant to apply
-
sampling frame
list of all possible elements in the population from which the sample will be selected
-
sample
subset of a population, elements selected for investigation
-
representative sample
- sample that is a microcosm of the population
- "represents" the essential characteristics of the population
-
probability sample
sample selected using random process so every element has equal chance of being selected
-
non-probability sample
- sample selected using non-random method
- some elements are more likely to be picked
-
sampling error
error of estimation that occurs when theres a difference between the characteristics of a sample and those of the population from which it was picked
-
non-response
when a unit selected to participate in the study refuses, can't be contacted, etc
-
census
attempt to collect data from all elements in a population rather than a sample
-
random sampling
- using random number generators or something to ensure that everyone has the same chance of getting picked
- chance and nothing else determines what elements are selected to be part of the sample
-
systematic sample
- selected directly from the sampling frame, ex. if 1/20 is the chance someone has to be picked, choose a random start within the sampling frame, then begin selecting every 20th person
- need to makes sure that there is no ordering/pattern within the sampling frame
-
stratified random sample
- stratifying the population into subgroups by a criteria (ex. faculty), and selecting a simple random sample or systematic sample from each of the resulting strata
- ensures sample is stratified in the same way as the population
-
multi-stage cluster sampling
- primary sampling unit isn't individual units of a population, but a cluster of them (aggregate)
- ex. sampling students from universities within specific geographic regions
- greater efficiency in data collection
- seek diversity in primary sampling units
- typically involve some kind of random sampling at the final stage
- sample weights may be needed!
-
_______ the size of the sample probably _______ the precision of the estimates it can create, and sampling error tends to _________
increasing, increases, decrease
-
response rate
% of the sample that actually participates in the study
-
non-response bias
- the extent to which people included in the sample differ from the population as a whole
- the people who don't answer the survey might be different somehow?
-
face validity
- the measure appears to reflect the content of the concept in question
- does it sound reasonable to you?
- assessed conceptually
-
content validity
- can be gauged by employing a criterion relevent to the concept in question bu on which cases differ
- does the item (or scale) cover the range of possible content?
- assessed conceptually, typically by experts in the field
-
convergent validity
- might be gauged by comparing it to measures of the same concept developed through other methods
- does a (new) measure of a concept correlate with other accepted measures of that concept?
- assessed empirically
- ex. student attendance rates and student satisfaction surveys
-
construct validity
- seeing whether concepts used in the research relate to each other in a way that is consistent with what their theories would predict
- is a (new) measure of a concept correlated with other variables that you predict (theoretically) should be related?
- assessed empirically
- ex. assignment - education and environmental concerns
-
internal validity
is the independent variable really affecting the dependent variable or is something else responsible?
-
spurious relationships
is the presumed cause (x) really responsible for the outcome (y) or is there a third variable affecting both?
-
things needed to prove causality?
- evidence of correlation
- temporal ordering (cause precedes effect)
- non-spurriousness (absence of alternative explanations for the correlation)
-
external validity
can you generalize from this study to other settings? the population?
-
sources of error in survey research? (6)
- sample bias
- random sampling error
- non-response error
- measurement (how you ask the question, reliability and validity of measurement)
- data preparation error
- interpretive error
-
confidence intervals
can statistically estimate the amount of random sampling error
-
sample weights
- fractions used to generate population estimates from disproportionate stratified random samples
- can also be calculated in some cases to handle different levels of non-response
-
what to do if you can't create a random sample?
- consider a non-probability design
- build it to make sure it is as representative as possible
- try to identify potential bias
-
voluntary samples
- volunteers are self-selected on the basis of interest in the subject
- most problematic with bias
- researcher has almost no control over how the sample is generated
-
convenience sample
- non-probability sample
- subjects are conveniently available to the researcher (ex. psych undergrads required research participation)
- not the best way to construct a representative sample
- external validity problems
-
quota/judgement/purposive samples
- non-probability
- used to find members of a highly specific population or build a reasonably representative sample of a larger population
- ex. going out and looking for people to interview in certain locations
-
snowball samples
- non-probability
- sample members provide names of additional potential sample members
- try to increase representativeness by starting several "snowballs"
- researcher has some control over sample composition
-
criteria for conducting effective surveys
- obtaining a representative sample
- obtaining reasonably high response rate
- obtaining valid and reliable responses to questions
- efficient (time and money) data collection
-
response sets
- sample members respond to a set of questions in a similar way, responses are being motived by something other than the questions being asked
- boredom
- social desireability
- acquiescent
-
acquiescent
trying to be agreeable by answering the same way (ex. agree, agree, agree)
-
reasonable goal (in today's age) for response rates?
over 50%
-
don dillman
- the "tailored design method"
- an extension of the social exchange theory
- explains why individuals are motivated to engage in certain social behaviors and not others
- need to install trust in participants, tell them how important their participation is, make questions interesting/engaging
-
guidelines for constructing survey questions
- establish researcher-participant relationship to increase participant engagement
- don't make participants do your work
- obtaining reliable and valid information
-
keys to professional-looking questionnaries?
- covers
- nice paper
- large enough font
- lots of white space
- return-address envelope
-
what makes an effective cover letter?
- establish legitimacy
- explain requirements of respondents (survey content, time, additional expectations)
- address ethical issues
- concise, professional
-
anonymity or confidentiality?
- anonymity: you don't know who the answers are coming from
- confidentiality: you won't tell anyone the info given by respondents
-
response effects
- factors that can lead to systematic measurement error
- ex. task effects, interviewer effects, respondent effects
-
halo effects
- bias due to missing response options
- bias due to unbalanced response scales
- bias due to provision of selective information (don't tell you as much information as you want)
-
pros of closed questions
- easy to process
- enhance comparability of answers
- some provide question clarification for respondents
- can answered quickly and easily
- reduce risk of interviewer/transcriber bias
-
cons of closed questions
- answers may lac sponteneity and authenticity
- must make sure answerable categories don't overlap
- can be difficult to make exhaustive
- respondents may differ in interpretations of forced-choice answers
-
pros of open-ended questions
- respondents can answer in their own terms
- allow unusual, maybe unexpected responses
- since the questions don't suggest answers, you can tap into a participant's knowledge and ideas about the issue
- can maybe generate fixed-choice answers
-
cons of open-ended questions
- time consuming to record
- answers have to be coded
- prospective respondents may put off having to write out a full answer
- may be inaccuracies in writing down exactly what respondents are saying
-
solutions for high rate of non-reponse with open-ended questions?
- better interviewers
- sharpened questions
- not too many open-ended Qs
-
what should you keep in mind when designing the questionnaire, observation schedule, and coding frame?
- the data analysis
- stats techniques used depend on how a variable is measured
- size and nature of a sample can limit the suitability of certain kinds of stats
-
levels of measurement
- nominal
- ordinal
- interval-ratio
-
nominal variable
- categorical
- composed of categories that have no relationship to each other except that they are different
-
ordinal variable
- values can be ranked
- ex. Likert-style questions
-
interval-ratio variable
- based on a unit of measurement
- takes the form of actual numbers
-
measures of central tendency for univariate analysis?
-
measures of dispersion for univariate analysis?
-
frequency table
- for univariate analysis
- provides the numbers and percentages of cases belonging to each of the categories of the variables in question
- can be created for all 3 variable types
-
contingency table
- for bivariate analysis
- like a frequency table, but allows two variables to be analyzed simultaneously to examine relationships between them
-
for cross-tabs, where should you total up the percentages?
percentage to 100% in the direction of the independent variable
-
bigger the sample, the _______ the confidence interval
- smaller
- significance is easier to obtain with bigger samples
-
interaction effect
ex. in multivariate analysis, when a statistical relationship exists for some groups but not for others
-
when determining significance in an experiment:
- set up a null hypothesis
- establish an acceptable level of statistical significance
- determine the statistical significance of the findings
- decide whether to reject or not reject the null hypothesis
-
goals of a structured interview
- ensure interviewees receive the same form of questioning and the same interview stimulus
- allow interviewees answers to be aggregated to form group rates
- means that the variation in answers is due to "true/real" variation in the characteristic being measured rather than extraneous factors
-
intra-interviewer variability
an interviewer isn't consistent with the way they ask questions or record answers
-
inter-interviewer variability
when there is more than one interviewer who may not be consistent with one another on how they ask Q's or record A's
-
coding frame
- rules for assigning answers to categories
- variation can occur in the way things are categorized bc of how the interview schedule was administered, or the way answers were recorded
-
questionnaire
- structured interview without an interviewer present
- make sure they are easy to follow
- less open-ended questions
- shorter to reduce the risk of "respondent fatigue"
-
reliability of a respondent throughout taking a survey is about ________
internal consistency
-
how to check internal consistency when constructing an index?
- Cronbach's alpha
- 1= perfect internal reliability
- 0= no internal reliability
- 0.80 usually used as an acceptable standard
-
nominal definition
describes in words what the concept means, ex. a dictionary defn
-
operational definition
- spells out the operations the researcher will perform to measure the concept
- ex. how to measure the incidence of crime?
|
|