reading 4 (p. 159-175) Flashcards
research design
specifies the sort of evidence you need to investigate your hypothesis, describes how evidence will be collected and analysed
research design vs method data collection
research design = overall strategy employed to demonstrate or investigate your argument
- what sort of evidence will test my hypothesis?
- I need this…
method of data collection = the sort of data that will enable you to implement this strategy, and where and how they will be obtained
- how and where can I get this information?
- I will get it through this…
attributes good research design
5
- specifies the type of research and techniques of data collection appropriate to the objectives of the project
- makes explicit the logic which enables you to draw inferences (logical conclusions based on evidence)
- identifies the type of evidence that confirms your hypothesis/argument, but is also a convincing TEST of your hypothesis (think about counterarguments and evidence that could proof the hypothesis wrong)
- decreases threats to the internal and external validity of your findings
- ensures that your findings are reliable
inference
= the process of using facts we know to draw logical conclusions about facts we don’t know
in research this process must follow certain rules : must be clear, orderly and systematic enough that someone can retrace your steps = transparent and explicit
internal vs external validity
- internal validity = the extent to which we can be confident that the independent (causal) variable produced the observed effect
*you need to rule out alternative explanations and other factors that play a role - external validity = the extent to which results from a study can be generalized beyond the particular study
e.g. laboratory experiments high internal validity, low external validity
e.g. field experiments higher external validity
(validity)
a study is valid if its measures actually measure what they claim to measure
reliability
the ‘repeatability’ or ‘consistency’ of your findings
reliable when other researchers can perform the same procedures and come with the exact same results
types of research design most commonly used in polsci
- experimental designs
- cross-sectional designs
- longitudinal designs
- comparative designs
experimental designs
- strengths, weaknesses
- 3 main ways of carrying out research
- widely regarded as being the most effective design for testing whether or not two variables are causally related (through rigorous use of experimental control = it rules out other potentially important variables)
- strengths = researcher can control the environment and manipulate variables with great precision
- defining characteristic = intervention by the researcher in the data-gathering process
3 main ways of carrying out experimental research =
- laboratory experiments: subjects are recruited + common location + researcher has most control possible over the environment = we can make certain conclusions about the influence of a variable
-
field experiments: intervention takes place in real-world, subjects may not now they are participating (ethical issues)
- strengths: random assignment makes unbiased inference + natural settings make sure it is useful info about the real world - natural experiments: naturally occurring events as interventions
! it is best to use methods together to combine external and internal validity (e.g. both labaratory and field experiments)
cross-sectional and longitudinal designs
- cross-sectional = analysis of a sample/cross-section of a population at a single point in time
- longitudinal = explores changes or trends over time (research is repeated with same sample or set of cases over two or more intervals) = useful for understanding dynamics of change + predicting long-term/cumulative effects
e.g. as countries get richer income distribution becomes more unequal ->
- cross-sectional: sample of countries with diff. eco growth + look at inequality
- longitudinal design: select a country or countries that have moved from lower to higher levels of development to look at income distribution over time
Two main types of longitudinal design =
- cohort study = selection group within a delineated population (e.g. age or ethnic group) + chart individual development processes
- panel study = collection of info about a sample (often randomly selected) collected at 1+ moments of time (often samples are not equal sizes over time because of attrition -> possible bias)
comparative designs
- most widely used researched design in polsci
- cases selected based on that they are: critical (testing a theory), revelatory (reveals relationship), unusual
large-N, small-N, and single-N:
- single-N/case study = used to uncover causal paths and mechanisms + assess specific mechanisms identified in theories (aka process tracing)
- large-N: to locate strong statistical relationship between two variables, might use theory to describe statistical results
- small-N = in-depth investigation of e.g. theoretically anomalous cases + existence of mechanisms
! case-study research is good with threats internal validity: it accounts for history/context and maturation (natural changes that affect the relationship between dependent and independent variable)
! case-study research is generally weaker in external validity: includes only a small number of cases (larger amount of cases = better generisable, BUT risks conceptual stretching)
historical design
- general + specific designs
- helps discover influence of certain events on others by contextualizing events (to better understand the relationship) + help understand how timing and sequence of social actions and processes impact on their nature and on the development of subsequent actions and events
different historical designs:
- historical events research = one case, one period of time (focus on event that has had an important impact on subsequent events or has important implications for a general theory)
- cross-sectional comparative research = study of many cases from one time period
- historical process research (longitudinal) = series of events within a single case to understand how processes change over time
- comparative historical research = series of events across many cases
- historical institutionalism = explore elements of persistence and change in institutions
- event structure analysis = probe links between actions that constitute the chronology of a given event for potential causal ties among those actions
‘test’ of a normative hypothesis
- logical consistency of the ideas
- consideration of the arguments of other theorists who have written about the issues
chief concern = present a clear, logical, rationally convincing argument
requires: tracking down persuasive arguments + philosophical methods as conceptual analysis, logic, classification
data-gathering strategies and data designs
explicit combinations?
NO: any type of data collection method can be used for any type of research design
boundaries for the study you must set before collecting data
- temporal domain (time frame)
- spatial domain (place)
- actors or units that are relevant
- variables or factors you think are important for answering your RQ + are components of your hypothesis
