Qualitative and Quantitative Research Methods Flashcards
1
Q
What is empirical vs normative research?
A
Empirical: study of what is
Normative: study of what ought to be
2
Q
What is the difference between positivism and interpretivism?
A
Positivists believe the social and natural worlds are the same - i.e. there is an objective reality that we can observe as scientific knowledge (law-like generalisations and cause-effect relationships can be discovered)
Interpretivists believe that the social and natural world are fundamentally different. Social phenomena are what we experience them to be. Scientific knowledge about the social world can only be gained through interpreting meanings. Direct observations not possible - it is all interpretation.
3
Q
What is ontology vs epistemology?
A
Ontology - the study of the nature of being, of existence
Epistemology - the study of how we can know about what is, it is the study of knowledge
4
Q
What are the two chief ‘types’ of research questions?
A
Descriptive questions - what are the characteristics of y? what is the distribution of y? changes of y over time/context? - about understanding the thing (not what causes it)
Explanatory questions - why does x happen? does x cause y? does z condition the effect of x on y? - about understand causes and and causal relationships
5
Q
What are 3 other types of questions?
A
Predictive questions - forecasting studies, aiming to understand/predict the future (weather, election outcomes, war outcomes…)
Prescriptive questions - how to bring about a policy/political outcome (e.g. how can we increase voter turnout?)
Normative questions - what is just or right or the best? (e.g. what are the properties of an ideal democracy?)
6
Q
What are the properties of a good research question (4 of them)?
A
-Researchable (sufficiently focused, no logical fallacies, avoid the common mistakes in questions…)
-Not yet definitively answered (e.g. are bananas yellow is a shit research question - it has already been answered - you want to make contributions to knowledge)
-Socially relevant - real world importance
-Scientifically relevant - should contribute to scholarly literature (e.g. academic debates or theories)
7
Q
What are common mistakes/logical fallacies in research questions?
A
-False premises/begging the question (e.g. why are women in the UK less likely to vote than men)
-Questions that cannot be answered using systematic research (e.g. whether something was inevitable)
-Tautologies - don’t say two things that have to be assessed different (e.g. are elections in China free and fair?)
-False dichotomies (e.g. do you support the death penalty to are you a soft-crime liberal)
8
Q
What are literature surveys and literature reviews (they are not the same)?
A
Literature survey - survey of what existing studies have asked and found, what are strengths/weaknesses of existing literature? - broader
Literature review - more comprehensive and critical analysis of existing research - synthesises key important findings and identifies gaps in knowledge
Both help to develop a research question in conversation with existing literature
9
Q
What are the three building blocks of theory (hint begins with concepts)?
A
-Concepts
-Their relations
-Conceptual units and levels
10
Q
What is theory in a broad sense?
A
An attempt to make sense of a complex world
-Description or explanation of an aspect of the world
-Generalisable - applicable to unstudied contexts, including the future
-Informed by empirical evidence
11
Q
What is theory in a narrower sense?
A
A theoretic answer to a research question
A set of assumptions about the world that in combination provide an answer to a RQ - informed by existing research
12
Q
What are concepts?
A
A collection or class of things that are to be regarded alike - it provides a label or general term to observations or events which are somehow alike
Conceptualisation is a lot about definitions!
Concepts need to be clearly defined - different people may understand the same concepts differently BUT a good definition gives the set of necessary and jointly sufficient conditions for a concept
13
Q
Why do concepts matter to variables?
A
We must form the concept of a variable before we can measure them.
Interesting concepts tend to be variables - empirical research requires variables
14
Q
Is establishing expected relationship between 2+ concepts a critical part of theory building?
A
Yes - a good theory establishes the expected nature of the relationship (i.e. as x increases, y decreases), provides a clear argument for the expected relationship (e.g. a causal mechanism) and builds on existing academic literature
15
Q
What is a causal relationship (simple)? And a causal mechanism?
A
One concept brings about change in another
The chain of events or conditions linking X to Y, how the causal effect is produced
16
Q
What are the dependent and independent variables?
A
Dependent variable: the outcome under study that one seeks to explain - the variable that is changed
Independent variable: variable that is thought to affect the outcome under study - the variable that causes change
17
Q
What are mediators in the causal mechanism?
A
Intervening variables (between independent and dependent) that transmit the effect (variables along the causal chain)
18
Q
What are the different levels of concepts? (like Econ)
A
Micro: individuals, citizens, voters, soldiers, etc.
Meso: political parties, interest associations, ethnic groups, rebel groups, etc.
Macro: countries, regions, economies, international system, etc.
Theoretical explanations can link different micro-level concepts… e.g. educational attainment links to political participation
Theoretical explanations can also link different meso/macro-level concepts… E.g., consensus democracy links to societal peace
19
Q
Do meso/macro phenomena often have micro-level foundations?
A
Yes - e.g. Coleman’s bathtub
Micro level things like wealth differences between ethnic groups (situational mechanisms) can lead to a perception of unfair treatment by the state (action-forming mechanisms) leading to anger etc and a feeling that the state needs to be changed radically (transformational mechanisms) WHICH in turn results in civil war (macro!)
Micro-level grievances in regards to wealth differences can be the foundation for a macro-level event - e.g. civil war
20
Q
What would be an example of the independent, mediating and dependent variables we could use in a study towards why some citizens put themselves forward for election?
A
Ind: Politicised Upbringing (or maybe historical exclusion of demographic group as an explainer for less candidacy)
Med: Political ambition (has an effect but not what we are measuring - and generally hard to measure objectively)
Dep: Candidacy
21
Q
Is theory a simple description or explanation of the world? An inference of the known? A way to make sense of the world?
A
Yes.
22
Q
Does theory often make more than just one empirical prediction?
A
Yes.
23
Q
Does a good theory need to be informed by empirical evidence?
A
Yes - a good theory is informed by empirical evidence (i.e. data)
24
Q
What is induction in theory-building and interpretive research? What are the stages?
A
Induction is the move from evidence (data) to theory.
The stages are…
-RQ
-Broad hunch
-Evidence (inductive research)
-Theory (inductive theory does not state a firm hypothesis)
25
What is one of the biggest flaw of induction?
The black swan problem - the data set used may be a complete anomaly
26
What is deduction in theory-building? What are the stages?
The move from theory to evidence.
The general idea of it is that theories, though possibly underpinned by a sound logical argument, need to be tested by data.
The stages are...
-Research question
-Theory
-Hypothesis
-Evidence
27
In deduction should falsification be prioritised over verification?
Yes - most argues that falsification should be prioritised over verification - the more good tests a theory survives, the more confident we can be
28
Is deduction now the standard approach in positivist-inspired political science?
Yes
29
What is the key difference between positivism vs interpretivism?
Positivism - Assumes reality exists
Interpretivism - Rejects fixed facts
30
Does deductive research require one or more hypotheses?
Yes - deductive research requires one or more hypotheses.
A hypothesis is a concise statement of an observable implication of a theory.
31
Must a deductive research hypothesis be falsifiable?
Yes - it must be falsifiable
32
Explanatory hypotheses at minimum must involve...?
-Independent variable/X/treatment (that which explains/affects the other)
-Dependent variable/Y/outcome (that which is explained)
-Relationship (e.g., +/-)
-Make clear whether probabilistic or deterministic
33
What can descriptive hypothesis be focused on?
The properties, characteristics, or distribution of a single concept
34
What is a research design? What 3 things must it have in regards to the sort of evidence, evidence collection, and evidence analysis?
A strategy for providing a test or investigation of a working hypothesis.
Must specify the sort of evidence needed to investigate the hypothesis
Must describe how the evidence will be collected
Must describe how the evidence will be analysed
35
What are 3 examples of possible research designs?
Experimental designs, comparative case studies, panel studies...
36
What are 3 examples of possible data collection methods?
Surveys, interviews, text analysis...
37
What are the two key key general things that every research design must do (something related to what you must do to a concept and units of analyses)?
-Operationalise key concepts
-Specify unit of analysis
38
What is the operationalisation of a concept?
The further specification of a definition of a concept to make the concept measurable - this is known as an operational definition - a specific definition allowing us to empirically measure a concept
E.g. the term 'state capacity' needs to be operationalised with a conceptual definition - e.g. a country's coercive state capability refers to a country's overall policing and military capabilities
Operational definition is often highly specific in quant research (can be looser in qual research)
39
What is a unity of analysis?
An entity being analysed, the lowest-level unit of an analysis (may or may not be of theoretical interest)
40
What are the three main levels of analysis?
Micro: individuals
Meso: parties, rebel groups, NGOs, ethnic groups, institutions, etc.
Macro: entire countries, societies, etc.
41
Does the unit of analysis follow neatly from the theory/conceptual unit?
Sometimes - for example, theories about individuals are often best tested with individual level data
42
Are conceptual units and units of analysis the same thing?
They can be - e.g. theories about voting behaviour are assessed using country - or region-level data. However, there are potential caveats, such as risk of ecological fallacy.
43
Ultimately, is the unit of analysis determined by the research question, theory, and what is possible/desirable empirically?
Yes
44
What are the stages of developing a unit of analysis in deductive research?
-Develop an ad-hoc theory
-Formulate a hypothesis
-Operationalise the independent variable
-Operationalise the dependent variable
-Find a good unit of analysis to test your theory (e.g. individual level theory needs an individual level of analysis)
45
What is another name for explanatory questions?
Causal questions
46
Why is answering causal questions using observational data tricky?
We cannot directly observe causal effects - it is the fundamental problem of casual inference
47
What is causal inference?
Inferring something we do not know (causal effects) from something we do know (data)
48
Does correlation equal causation?
NO!
There may be an association between two variables (which is a necessary factor for establishing causation) BUT this is not the only requirement for causation
49
What are two key requirements for establishing causality?
1 - Ruling out all confounders (confounding variables)
2 - Ruling out reverse causality
50
What is the idea of the problem of confounding variables in explanatory research?
Confounders are third variables which are related to both X and Y providing an alternative explanation for an association between X and Y
E.g. Chocolate consumption (X) and No. of Nobel Laureates (Y) are associated - but the confounding variable is that the countries with more chocolate consumption tend to be richer and so more laureates.
So there is an association between X and Y, but there is a confounding variable of wealth associated with both variables BUT X and Y are not related in a causal mechanism - there must not be any confounding variables to confidently establish a causal relationship between X and Y
51
What is the idea of the problem of reverse causality in explanatory research?
I.e. the dependent variable causes independent variable to change, not the other way round
Need to make sure that the dependent variable is not mistakenly affecting the independent variable when in research it should be the other way round.
52
What are randomised experiments (form of experimental research design)?
A research design in which the researcher both controls and randomly assigns values of the independent variable to participants - increasingly used in social sciences
53
What is a distinguishing feature in experimental research design vs observational research?
Key change: Researcher intervenes in the data-generating process (DGP)
Researcher assigns different treatments to study participants and the researcher assigns those treatments randomly
54
Are experiments the gold standard for causal relationships?
Yes
55
What does random assignment in experimental research design ensure?
Control groups are comparable based on their pretreatment characteristics - includes any known confounder (and even unknown confounders) BUT requires a significant population pool for research
Furthermore, with the treatment preceding the outcome reverse causality is also ruled out
56
What methods can be used to identify causal effect in experimental research?
Using statistical methods, such as two-sample t-test, linear regression, etc...
57
What is internal validity?
The degree to which we can be confident that a study identifies the causal effect of the independent on the dependent variable
58
What is external validity?
The degree to which findings can be generalised to other contexts
59
Do experiments with random assignment tend to fare well in terms of internal validity?
Yes
60
What are the 3 reasons for potential weaknesses in random experiments in terms of external validity?
Reactivity: People may change their behaviour when they know they are being observed
Population validity: Experiments often involve unrepresentative subject pools (e.g., UG students) and it can therefore be questionable whether experimental findings generalise from the study sample to the population of interest
Ecological validity: Behaviour observed in artificial experimental settings may not generalise well to the real world
61
What are the three main types of experiments?
Laboratory experiments (subjects recruited to a common location)
Field experiments (carried out in natural environments)
Survey experiments (conducted in the context of a survey)
62
Why is not all research experimental? Why does some of it have to be observational?
Practical limits/objections: Many things political scientists are interested in are hard or even impossible to manipulate
Ethical issues: subjects should not be harmed, deception often seen as problematic, etc.
External validity concerns
Observational research is. better for descriptive work (including interpretive work): experiments are for causal questions only
63
What is observational research design?
Research design where the researcher does not have control over values of the independent variable. The researcher does not intervene in the data-generating process - there is natural variation in the population.
Despite growth of experimental research, observational research is still the dominant method in social science
63
What is the typical/best use of observational research designs?
Description
64
What are observational research questions most commonly focused on (2 chief focuses)?
-Distribution (how democratic are different countries across the world? - can be over geography or time)
-Characteristics and meanings (what do elites mean by 'merit'?)
65
Why may observational research designs be weaker in answering explanatory questions (i.e. establishing causality)?
Treatments are not randomly assigned, and unconfoundedness and no reverse causality assumptions may be less credible
66
What are 4 possible ways of tackling confounders in observational studies?
Statistical control
Most-similar and most-different designs
Observing the same units over time (within changes)
Difference-in-differences, regression discontinuity, and other modern causal inference designs
67
Is there significant variability in terms of internal validity in observational research studies?
Yes - causal identification cannot be as strong as in a randomised experiment - most importantly since it is extremely hard to rule out all possible confounders
68
What are natural experiments?
Experiments where the values of the independent variable arise naturally in such a way that we can speak of true or, more realistically, “as if” random assignment - e.g. lotteries of any sort (Vietnam draft), random assignment of election observers...
69
What is data?
Data is the record of collected information - it is used as empirical evidence to indicate concepts.
70
Is empirical research possible without data?
No
71
In large C research what is the typical expression of data (quantitative or qualitative)?
Quantitative - numbers and scientific fact is analysed -numbers and statistics
72
Is quantitative data in large-c research highly standardised?
Yes data is highly standardised
73
In interpretivist small C research what is the typical expression of data (quantitative or qualitative)?
Qualitative - meaning is analysed - words and symbols
74
Is qualitative data in interpretivist small-c research highly standardised?
No - it is not really relevant to qualitative data of that nature
75
In scientific-realist small C research what is the typical expression of data (quantitative or qualitative)?
Both! - numbers and meaning are analysed - numbers, statistics, processes, words and symbols...
Middle of the road standardisation
76
What is an example of data we may analyse in large-C research vs in interpretivist small-C research? Bonus what would be an example of realist small-C research?
Large-C = Data Matrixes
VS
Interpretivist Small-C = Discourse analysis
Realist could be something like secondary analysis of an academic journal
77
What are the four key types of ordering of data (not relevant for interpretivist as data doesn't sit on measurable axes)?
Nominal - data classified into categories without a natural order (e.g. the categorisation of party affiliations into Conservative, Labour, Lib Dem)
Ordinal - data is arranged in a meaningful order BUT intervals between rankings may not be equal (e.g. highest level of education... high school vs undergraduate vs masters vs PHD)
Interval - data is on numeric scales with equal intervals between values but no true zero point (e.g. place yourself on a 0-10 scale of Liberal to Conservative where 0 is very liberal and 10 is very conservative)
Ratio - data is on numeric scales with equal intervals and a meaningful zero point (e.g. government spending in GBP - zero is a true zero and 100 million is 100 million)
78
What is operationalisation in research?
Operationalisation is going from concepts to actual measurement... the central process in data-focused research
You take a concept, then reach an operational definition and then measure it.
79
What is an operational definition?
Operational definition = the choice of observable indicators used as proxies for concepts - you then must reach a description of the process of how the concept will be observed (indicated)
Choice of observable indicators is key!
80
What is a measure in data-focused political research?
Specific values/categories of the operationalised concepts and a way how they assign to the measurement units (each unit of measurement examined and unambiguously assigned a value/category)
81
What are the three core steps in research?
Begin with a CONCEPT (e.g. Corruption)
Create an OPERATIONAL DEFINITION (e.g. indicators of perceived corruption include bribery, diversion of public funds or the ability of gov'ts to contain corruption)
Then using the observable indicators stated in the definition MEASURE these indicators (e.g. can measure the perceived level of public sector corruption on a scale of 0-10 in an expert survey?)
82
What is the question of validity in political research?
The question of validity is around how well does our/the measure indicate the concept!
How reflective is our measure of the concept we are trying to greater understand...
82
What are the four core types of validity (regarding measurable indicators in research)?
Face validity - is the indicator directly relevant to the concept?
Content validity - do the indicators cover the full range of the concept?
Construct validity - see on other flashcard - how valid is the indicator in relation to other indicators...
Criterion validity - see on other flashcard
83
What are the two types of construct validity?
Convergent validity - is the indicator similar to other indicators it theoretically should be similar to? - is it similar enough
Discriminant validity - is the indicator different from other indicators it theoretically different to? - is it different enough
83
What are the two types of criterion validity?
Predictive validity - can the indicator actively predict an outcome based on a concept it is meant to reflect? (e.g. can your attitude survey actually predict attitudes that will be acted upon_
Concurrent validity - does the measure correlate with another measure of a concept
84
What is the difference between reliability and validity?
Validity is how well does the measure indicate the concept
Reliability on the other hands is how accurate is the measurement
85
What are three ways you can tell if your measurement system is reliable (unified, sources, repetition)?
-Measurement is systematically unified across cases
-If multiple sources, opinions or codes agree
-If you repeat the measurement of the same unit, do you reach similar conclusions
86
Can a non-valid measure be reliable? Can a non-reliable measure be valid?
Non-valid measure can be reliable
Non-reliable measure however cannot be valid
As such, validity is someway an extension of reliability...
87
What are the two main types of data (two types determined by collection methodology)?
Primary data - (you) the researcher carries out the data collection
Secondary data - others have collected the data (you) the researcher is just collecting it
88
What are the pros of primary data?
The researcher has full control over the collection process
89
What are the cons of primary data?
More expensive and more difficult (requires a lot of effort to collect all the data)
90
What are the pros of secondary data?
Less expensive and faster, good for qualitative case studies
91
What are the cons of secondary data?
No control over the collection process (may be flaws), and you need to thoroughly check the quality of research
92
What would be the large C vs interpretivist small C version of collecting interview data?
Large C: Results from a survey
I Small C: Coded transcripts from interviews and focus groups
93
What would be the large C vs interpretivist small C version of collecting observational data?
Large C: Roll call votes from parliament
I Small C: Analysed field notes from participation observation (i.e. maybe transcripts from parliamentary debates)
94
What would be the large C vs interpretivist small C version of collecting document-based data?
Large C: Standardised codes of policy positions of parties from party manifestos
I Small C: Coded party manifestos (in-depth policy discussions)
95
What would be the large C vs interpretivist small C version of collecting secondary source data?
Large C: Gov't statistics
I Small C: Academic articles
96
What is a case?
A case is the subject of a study, the entity or entity type that is being examined in depth.
A spatially and temporally delimited phenomenon of theoretical interest
97
What is an observation?
The lowest-level unit in an analysis at which a measured variable can only take one value (unit of analysis)
98
What is a sample in research?
A set of cases or observations that are analysed in a given piece of research
99
What is a population in research?
A cases which in combination make up the universe of cases, i.e. all cases
100
If empirical research wants to say something about a population of cases, but is unable to study entire populations of cases, what do we analyse instead?
Samples of cases (i.e. samples from the wider population)
101
What is the risk/danger of sampling when sampling a selection of cases from a population?
The insights from the samples may not generalise well to the population of interest and there is also the added risk of sampling bias
102
What is survivorship bias case selection?
Case selection based on the dependent variable - e.g. survival... we want to explain survival and so we select only cases (lets say soldiers in a war) that survived
When selecting cases on the values of the dependent variable we risk making biased causal inferences - we should avoid such case selection in large-C research
103
Are there exceptions in which we can select cases based on the dependent variable?
Yes there are such circumstances - in particular, exceptions emerge in qualitative/small C research
104
Other than survivorship bias what are a few other circumstances that can lead to sampling bias?
Distorted sampling frames, selection on the dependent variable, survivorship bias, non-response bias, self-selection into the sample... avoid cherry picking cases that are known to support your hypothesis
Can't always be avoided but research studies should be open about it and consider how it may affect conclusions
105
Should you use the same set of cases you used in inductive research to reach a conclusion in subsequent testing of the resultant theory and hypothesis?
No - definitely not
106
Reminder, what is a large-C study?
A less intensive study of a large number of cases using quantitative methods
Both small and large C may use numbers, but in large-C research, attention is not on individual cases, but samples of cases. Unlike small-C research, large-C research necessarily relies on formal statistical analysis of matrix observations.
107
What is a data matrix?
A rectangular dataset displaying observations between variables
107
On a data matrix in large-C research what do rows display and what do columns display?
Rows = observations (unit of analysis - e.g. countries)
Columns = variables (i.e. a quant measurement of a concept - e.g. economic growth)
108
What happens to rows in a data matrix when temporal variation is added into the mix?
Before each row was a case (a country) - which before was the smallest unit of analysis - now each row represents a case in a time (e.g. a country in a year) and each row is a case with a time specification
E.g. before may have been Libya, Lebanon, Greece going down, now it is Libya 2022, Libya 2023, Lebanon 2022, Lebanon 2023, Greece 2022, Greece 2023...
109
What are large-C studies and quantitative research of a similar nature used for?
-Analysis of explanatory RQs using experimental data
-Analysis of explanatory and descriptive RQs using observational data
110
Is large-C research typically deductive or inductive?
Typically deductive - key reason is quantitative methodology requires a clear ex-ante idea of what one intends to measure. But quasi-inductive designs are possible…
111
What are the strengths of large-C studies?
-Increased potential for generalisability (facilitates the study of random samples from a population or even the entire population which are more generalisable)
-Increased ability to identify causal effects (statistical control allows controlling for confounders in observational studies and experimental research design of large-C also helps test causal effects)
112
What are the weaknesses of large-C studies?
-Less attention to context of cases, pressing concepts into numbers creates potential for reductionism and over-simplification, and there are no intensive studies of individual cases
-Less useful for inductive research (generally)
-Limited usefulness for interpretivist research due to the 'thin' form of analysis that is not very useful for detailed analyses of motivations and meanings behind actions
113
What is the twin goal in large-C research case selection (regarding sample size and representation)?
-Sample large enough to enable robust statistical inferences
-Sample representative of population
114
Which sampling strategies have the potential to achieve the twin goal of large-C case selection?
-Probability sampling
-Total population sampling
115
What is a limit to probability sampling?
Can be expensive or impractical - as such non-prob strategies can be an alternative (though often less representative)
116
What is total population sampling? What is an example?
Selecting all cases in a population - e.g. a census relatively common in comparative politics, political economy, IR...
Total population sampling could include a study of all countries in the world, the political standing of all ethnic groups in a country, all political parties in a system... sample size is not necessarily that large though
117
What is a key benefit in total population sampling?
High external validity and full representativeness
118
What is simple random sampling? What is the law of large numbers
The simplest form of probability sampling is simple random sampling - All cases are drawn from the population with the same probability.
Random sampling leverages the law of large numbers which refers to the idea that sufficiently large probability samples are therefore likely representative - larger the sample size the more it approximates the wider population
119
What is stratified random sampling?
Another common form of probability sampling - populations are divided/stratified into relevant subclasses - such as different genders, age groups, ethnicities...
Cases are then drawn at random from different subclasses ensuring that important groups are adequately represented - achieves more representative samples with fewer observations
120
What are some other relatively common types of random sampling (other than random or stratified random)?
Cluster random sampling
Multi-stage cluster random sampling
Systematic random sampling
121
Why would non-probability sampling be preferable to probability sampling?
Prob can be expensive and/or impractical - non-prob is frequently cheaper or more easily feasible
However insights from non-prob sampling are not easily generalised to the population of interest (explanatory research can give rise to sample selection bias)
122
What are some common types of non-probability sampling? Convenience sampling methods...
Volunteer sampling (e.g. surveys on news websites, participation requests on social media)
Snowball sampling
Quota sampling
123
What is a small-C study?
An intensive study of a single case or a small number of cases (e.g. single or comparative case studies)
124
What type of data (qual or quant) does small-C research typically use?
Qualitative research methods - e.g. content analysis, discourse analysis... and qualitative data - e.g. interviews, texts...
125
Is an observation effectively synonymous with a case?
No - observation does not equal case.
Small-C studies can have a large number of observations (e.g. in-depth interviews or archival sources with many data points)
BUT, small-C studies by definition look at a small number of cases
126
Is small-C research useful for descriptive and explanatory research questions?
Yes - can be useful for both
127
What are 2 potential limitations of small-C research (generalisability and confounders)?
Small number of cases hampers the ability to generalise from small-C research findings
Small number of cases adds extra complication for dealing with confounders
128
What can small-C studies do that large-C studies cannot?
Rich, in-depth, "thick" forms of analysis...
Quantitative measures can lack nuance and in-depth studies of cases can enable richer measurement of difficult concepts enabling the uncovering of hidden meanings in concepts
129
What is small-C descriptive inductive research?
Intensive study of small set of cases may reveal new explanations not considered by previous research to help form new conceptual/theoretical understandings
130
What is a 'thick description' in small-C research?
Detailed analysis of small number of cases enables putting human behaviour into context
Helps establish individuals’ motivations for action & their interpretations of the contexts driving their actions
131
What is the main case selection method employed in small-C research?
Purposeful sampling - strategic case selection with research goal in mind with the aim of maximising information gain
Requires careful thinking about the objective of a study, the nature of cases, and how they relate to the population
There are many different strategies with no general rules/prescription - it depends what makes sense in the context of the research
132
What are the two most common case selection in small-C descriptive studies?
Typical case(s) and diverse cases
133
What is the idea of the 'typical case' in small-C research?
Typical case = most representative case (the archetypal case) - a case which represents a larger population well on important features
Hope is that descriptive findings in typical cases will generalise to the population - common strategy when the goal is description
134
What is the idea of 'diverse cases' in small-C research?
Deals with the weaknesses of the typical case technique - populations are often diverse, the typical case may not exist
A descriptive study may also focus on several cases that, in combination, capture the diversity of a population...
Diverse cases selection aims to study a set of diverse cases with one defining feature in common (e.g. democratic countries)
135
What are common case selection technique in small-C explanatory studies?
Extreme cases, deviant cases, most-similar cases, most-different cases...
136
What is small-C explanatory inductive research?
Likewise an intensive study of small set of cases searching for new explanations for phenomena
137
For testing causal hypotheses in small-C explanatory research what are the most common case selection techniques?
Crucial cases, most-similar cases, most-different cases...
138
What is most-similar case design (small-C research)?
Min of two cases which are similar in terms of background conditions (Z) BUT differ in terms of X or Y
Can be used for exploration: identify the varying X that explains variation in Y
Can be used for testing: assess whether variation in X leads to variation in Y
139
For studying mechanisms (mechanism studies) in small-C research what is the most common case selection technique?
Pathway cases
Mechanism studies is the exploration/testing of causal mechanisms (what connects X and Y)
140
What is most-different case design (small-C research)?
Minimum of two cases which differ in terms of background conditions (Z), but are similar in terms of either X or Y
Can be used for exploration: identify the common X that explains common Y
Can be used for testing: assess whether common X leads to common Y
141
What is the idea of extreme cases in small-C research case-selection?
Involves looking at unusual values in Y - they are case studies of unusual phenomena - e.g. very high economic growth, genocide, revolution, world war...
Employed to try and find explanations for unusual cases (inductive search for new explanation)
142
What is the idea of deviant cases in small-C research case-selection?
Involves looking at one or more cases which deviate from a common causal pattern - an anomalous case that is poorly explained
E.g. the dog that didn't bark or the cat that did bark
The goal is the explanation of an anomaly (inductive search for new explanation - move from data to theory)
143
What is the idea of crucial cases in small-C research case-selection?
Involves one or more cases which have a profound impact o our confidence in a hypothesis
Most-likely case: an easy test that if it fails casts strong doubt on a theory. If theory does not even work in this favourable context, it is unlikely to work anywhere...
Least-likely case: a hard test that if it passes provides strong evidence for a theory. If theory works even in this most unfavourable context, it is likely to work everywhere...
Used for testing of causal hypotheses or establishing the scope conditions of a theory
144
What is the idea of pathway cases in small-C research case-selection?
The most common case selection method for exploring or testing causal mechanisms
Assumption: we have good evidence that X and Y are causally related
What explains the causal relationship? What processes link X to Y?... Select one or multiple cases in which both X and Y are present
Use process tracing to establish the causal process leading from X to Y
145
What are two more considerations to think about in small-C case selection (importance and logistics)?
-The intrinsic importance of cases
-The logistical aspects of cases - i.e. do you have access to archives, interesting interview partners, is there a language barrier?...
146
What is a survey?
A method of gathering data from large samples of individuals through their responses to a structured set of questions - data is collected through asking people questions
147
What is a large-C method?
Large number of respondents - often standardised response options
148
What are surveys good for?
Versatile - often used for description (what do people think about X) or explanation (why do people think the way they do about X)
Often used in observational research - but survey experiments are increasingly common (can be a primary or secondary data source)
149
What are surveys not good for?
-Limited to things that can be found out by asking people (people's attitudes, emotions, or behaviour need to be of theoretical interest)
-Surveys only respond to the present and not the past (can't survey and collect data on the victorians or people from the future)
-Often less useful for inductive (theory-generating) research (questions and responses need to be defined before data collection)
-Lacks ability to create intensive studies of individuals' motivations or the meanings they attach to their actions (it is a large-C method)
150
What is the ideal survey design?
The ideal survey provides accurate information about people’s attitudes and behaviours not just in the sample, but in the target population
In practice, not all surveys live up to this goal… Sampling bias or measurement errors
It is important to think carefully about survey design… Limitations are common BUT they should be clearly communicated and you should be aware of the issues…
151
Why do we have to sample in surveys?
Survey research typically wants to say something about populations of people (e.g. all eligible voter... or all 18-25 year old British men)
BUT... logistically we are rarely able to survey the entire populations (except via a census), instead, we usually have to draw samples of people from the population... whilst making sure that it represents the population
152
What has to be known about a whole population in probability sampling?
All members of the population have to have a known, nonzero chance of being selected into the sample
153
What is the law of large numbers?
With random selection, sample means tend to increasingly approximate population means as the sample size increases - samples likely to be representative
154
What is a simple random sample?
Respondents randomly selected from a larger group
155
What is a stratified random sample?
Respondents are split into sub-groups and then randomly selected from each group
156
Why might you not be able to do probability sampling?
Can be expensive, impractical, and are increasingly unachievable due to low response rates
157
What are non-probability samples?
Probability of selection of units into the sample is unknown... population inference more complex, and less certain!
158
What are the three main types of non-probability samples?
Quota sampling
Volunteer sampling
Snowball sampling
159
What is quota sampling in survey sampling?
Non-random selection of a predetermined number or proportion of units used to mimic the characteristics of a market/population - i.e. 20% of target population are males above 40 so you create quota of 20% for males 40+
Extremely uncommon now
160
What is volunteer sampling in survey sampling?
Participants who choose to participate in a survey participate
The population is only made up of the willing - so will be distorted in some manner
161
What is snowball sampling in survey sampling?
Existing study subjects recruit future subjects from their own social networks - therefore, the sample group is said to grow like a rolling snowball
162
Why may surveys not be representative?
-Small sample sizes
-Coverage error (segments of a population are not part of the sampling frame - maybe not everyone is listed on the electoral roll or in a phone book when sampling from these sources...)
-Incomplete population lists
-Drop off and item non-response (people only answer half the survey)
-Unit non-response
-Ideal survey is probability based but most are not
163
Are most survey questions closed or open ended?
Closed - there are standardised responses
Open-ended questions are possible, but make analysis much more complex (and answers may not be relevant)
164
What are the two main nominal question formats?
-Binary - e.g. yes or no
-Multiple choice
165
What are the four main ordinal question formats?
-Linkert scales (aka items) - completely disagree, tend to disagree, neutral...
-Ordered rating scales - e.g. very important to not at all important...
-Behavioural frequency - e.g. never, rarely, sometimes...
-Ranking - e.g. which party would you say you align yourself the most with conservatives, labour, lib dems and which would be the next most important
166
What is the main quantitative question format?
On a scale of 1 to 10 or 1 to 100...
167
Why do some survey questions include a 'don't know' or 'prefer not to say' option?
It is effectively a tradeoff between losing out on a response, but it also stops people being forced to make up answers
168
A good survey question is valid and reliable - what does this mean?
Valid: question measures what it is supposed to measure
Reliable: question provides consistent measurement
169
What are some good practices in creating good survey questions?
-Simple language (avoid technical jargon - needs to be accessible)
-Precise language (avoid ambiguity)
-Ask one question at a time (avoid double-barreling)
-Ask questions openly (avoid leading questions)
-Vary direction of questions (agreement should not always indicate a leftist position)
-Keep surveys as short a possible (to avoid straightening)
-Do not just ask people your explanatory research question
-Avoid social desirability bias (do not overreport good behaviour)
170
What is a survey experiment?
A randomised experiment conducted within a survey - there are a minimum of two experimental conditions and some survey respondents are exposed to one conditions, others to the other condition(s) at random
Due to random assignment, the researcher can assume that the only difference between conditions is the difference between experimental conditions (in sufficiently large samples)
171
What are the two types of survey experiments (i.e. what do they intend to do)?
-To identify or counter measurement bias
-To establish causal effects
172
What are list experiments in regards to surveys?
Unobtrusively measure sensitive attitudes (e.g. racial bias) in a way that allows respondents to remain truthful but also not explicitly reveal their sensitive attitudes
Present respondents with a list of times, and ask them how many apply to them. In one group (the treatment group) you include a sensitive item (e.g. attitudes towards black people) and in the other group - the control group - you do not include this item.
The avg difference between the treatment and control condition represents the percentage of respondents for whom the treatment item applies - i.e. you learn about the prevalence of the sensitive item in the population
173
How could you implement experimental manipulation into surveys to establish causal effects?
-Hypothetical scenarios
-Argumentative frames
-Textual or visual primes making a topic more salient in people's mind
-Encouragement to perform certain tasks
174
Why would you implement experimental manipulations into surveys?
To establish causal effects
175
What is ontology?
What's out there to know and what is the nature of existence?
176
What are the two main ontological approaches we study?
Objectivism - the world exists independently of researchers (related to positivist epistemology)
Constructivism - researchers have a role in creating the world and human knowledge (related to interpretivist epistemology)
177
Interviews and focus groups are an example of qu____ research?
Qualitative research - most often small-C research
178
What are interviews in political research?
Conversations with a purpose - generally one on one and normally recorded/extensively noted
179
What are focus groups in political research?
Involves questions to multiple respondents (4-10) with a moderator/facilitator
Usually focused on a single topic or issue (possibly more controversial or difficult to discuss one on one) - respondents may challenge each other during sessions (provides more realistic response and in depth understanding)
Often recorded
180
What would be the positivist explanation for why a researcher may do interviews or focus groups?
To find out specific facts or information from participants
181
What would be the interpretivist explanation for why a researcher may do interviews or focus groups?
To see how the interviewee/subject explains things in their own terms and how they interpret the world
182
What are interviews and focus groups good for?
Providing first-hand data that we cannot get anywhere else (first-hand knowledge of an institution/policy area/experience)
Providing data on personal experience and private emotions (window into an array of human experiences and help us to understand how people make meaning)
Helps to see links between micro and macro (interviews focus on personal experience - help us to understand link between structure and agency)
183
What are interviews and focus groups more useful for - theory building or theory testing?
Theory building rather - helping to understand the world and identify patterns to develop theory
184
What are interviews and focus groups not good for?
Large N research (not always though)
Reliability - interview participants may lie or say what they think the interviewer wants to hear not what they believe or is true
Accessing hard to reach groups (e.g. senior politicians)
Theory testing (tends to be more exploratory - especially in semi-structured interviews)
185
What kind of research questions can we ask with interviews and focus groups (empirical understanding, meaning, micro/macro links)?
Solving an empirical puzzle (WHY do parents choose not to vaccinate their children?)
Questions about meaning (what does the feminist in feminist foreign policy MEAN?)
Questions about links between micro and macro level (can help to recreate decision-making processes - open the 'black box' of the state - e.g. why do anti-gender governments engage in pro-gender equality measures?)
186
What kind of research questions can we not answer with interviews and focus groups?
Causal questions - not great for causal mechanisms (e.g. did economic uncertainty case Brexit)
Questions that require a large dataset to fully understand (e.g. why did Scotland vote against independence in 2014?)
Questions about the past (people are unreliable and forget - different if we are interested in questions about people's perceptions of the past)
Theory testing questions (may make interviews too structured/reductive)
187
What are the three 'structure-levels' in interview types?
Highly structured (closed) - questions and answers are pre-coded
Semi-structured - some questions agreed to set agenda
Unstructured (open) - conversational: may be more personal
188
What are some ethical considerations that must be made when making interviews and focus groups?
-Do you need to ensure anonymity? Is informed consent possible?
-Do you have permission to record?
189
How do positivists do interviews different to interpretivist?
To positivists setting/broader context is unimportant - it is the content of what is said and the facts
To interpretivists setting/minor details/context are important and telling
190
Will who you are as a researcher have an impact or even shape the interview?
Yes - e.g. women may have an advantage in interviewing men on difficult/controversial topics as they are perceived to be of little threat a disadvantage - can lead to an assumption of shared understandings which may inhibit critical analysis!
191
Can you use data analysis on interviews/focus groups? (matrices and networks?)
Yes
Data reduction - process the interview into collected data - e.g. turn interviews into transcripts then begin the coding process
You could then put the data into matrices (tables to identify patterns) or networks (identify nodes/concepts/policy developments and the connections between them) to help develop conclusions
192
What are some problems with interviews (and probably a lot of qualitative research)?
The main one is - is it overly subjective?
Can we infer anything from interviews... and will this type or research really be enough to inform policy (if we have a normative approach)?... are other considerations to make
193
What are the three main techniques utilised under the umbrella of 'textual analysis'?
Content analysis
Critical discourse analysis
Rhetorical analysis
194
What is textual analysis?
A way of analysing written, verbal or visual communication messages - involves understanding language, symbols, and/or pictures present in texts to gain information regarding how people make sense of and communicate life and life experiences
195
Is textual analysis nearly always interpretive?
Yes - it is focused on understanding how each researcher, audience, or viewer brings understandings of the world shaping interpretations from the text - it is not so much about the factual content of the text but about how others understand the world
Positionality of the researcher (like in interviews) therefore plays a key role
196
Is textual analysis useful for both descriptive and explanatory research?
Yes
Description - useful for identifying what information is present in social/political documents - e.g. what does party x think about immigration?
Explanation - going beyond description to interpret and analyse underlying meanings, relationships, and implications, with a focus on human behaviour through communication - e.g. how does a politician construct/frame immigration as a problem
Great for analysing secondary data sources
197
What is textual analysis less useful for? (quant, causation...)
Statistical analysis
Determining causal relationships
Quantifying data
198
What is content analysis?
An analysis of the content of communications to interpret meanings the meanings within them - a careful detailed examination and interpretation of a particular body of material in an effort to identify patterns, themes, biases, and meanings
199
What are the three major stages in the content analysis process?
Preparation - select sample of the data
Organising - decide on unit of analysis (focus on words, sentences, entire corpus of text...) and code the data (identify themes to attribute to units of analysis)
Reporting - what patterns have you noticed in the data?
200
What is coding in the organisation of text in content analysis?
Coding = the condensing of qual data into smaller analysable units through the creation of categories and concepts derived from the data
Condensing/distilling words into fewer content-related categories basically
There are two types of coding?
201
What is deductive coding? What is inductive coding?
Deductive coding - researchers work from pre-existing themes derived from existing theories and research - you draft categories BEFORE you start analysis!
Inductive coding - researchers immerse themselves in documents to identify different themes - you allow categories to emerge WHILST you analyse texts
202
What is the difference between manifest and latent content analysis?
Manifest - focuses on the surface level and directly observable
Latent - goes beyond the surface to interpret the underlying/implicit/symbolic
203
What is the difference between conceptual and relational reporting in content analysis?
Conceptual - determines how often a concept appears but does not examine the relationships between concepts (quant approach)
Relational - goes beyond just identification of concepts and examines how they are related within a text (qual approach)
204
What are three advantages of content analysis?
Unobtrusive data collection - no direct involvement of the participants - presence as a researcher doesn't influence results
Transparent and replicable - systematic procedure - yields results with high reliability
Highly flexible - can conduct content analysis at any time, location, no costs...
205
What are three disadvantages of content analysis?
Reductive - focusing on words or phrases in isolation can sometimes be overly reductive, disregarding context, nuance, and ambiguous meanings
Subjective - affects reliability and validity
Time intensive - manually coding large volumes of text is extremely time-consuming, and it can be difficult to automate effectively
206
What is critical discourse analysis?
A qualitative analytical approach for critically describing, interpreting, and explaining the ways in which discourse construct, maintain, and legitimise social inequalities
It is critical int eh sense that it aims to contribute towards a fairer and more just society
207
How does critical discourse analysis view discourse?
Language use in speech and writing - it is seen as a form of social practice which both shapes and is shaped by events
208
Is power and hegemony central to CDA?
Yes - CDA views language not just as a tool for communication BUT also a means to enact and maintain power relations
Key to CDA is the concept of power, or the chance that a person in a social relationship can achieve their own will against the will of others
209
What is the ideological square in CDA?
The idea of the discursive construction of 'us' and 'them'
Centres on an emphasis of 'our' good things and 'their' bad things whilst de-emphasising our good and their bad - general premise is positive self-representation and negative other-representation
Central to discourses of identity and difference, such discourses are salient for discourses of discrimination
Square shows us the discursive reproduction of the idea of positive 'us' and negative 'them'
210
Does CDA often focus on linguistic concepts?
Yes - CDA often targets common linguistic concepts such as time, tense, modality, actors, and argumentation, word order, coherence, intonation, topic choice, turn-taking, hesitations, pauses, and laughter and voice, and choice of words
211
What is argumentation in regards to CDA?
Argumentation is a discursive strategy that relates arguments that are employed to substantiate the positions in a discourse, using typical topos which which 'justify the transition from argument to conclusion'
Argumentation focuses on the use of this core concept of a topos - a 'taken for granted' argument - e.g. the topos of reponsibility - it is clear in political messaging that political movements use the taken for granted argument of responsibility over controlling our borders as seen by the utilisation of this talking point on both sides of the Brexit debate
The assumed nature of the topos simplify the movement from an argument tp conclusion and are used to justify arguments and depict them as 'rational' and therefore beyond debate
212
In CDA what are the three dimensions of every instance of the use of language?
The actual text - speech, writing... the linguistic features of the text
The discursive practice involving the production, consumption and distribution of text - interpretation of the discursive practice is important
The social practice - the social and historical context of the use of language
213
What is Wodak's idea of the discourse historical approach to textual analysis?
An approach to textual analysis that emphasises the contexts of discourse int he process of explanation and interpretation of texts on social issues such as sexism, racism and anti-semitism
214
What are the three stages (analytical framework) of the discourse historical approach to textual analysis?
Stage 1- identify contents or topics (initial coding of prominent themes)
Stage 2 - investigate discursive strategies (in-depth analysis of strategies)
Stage 3 - explore the textual features and techniques (micro level analysis of linguistic means/features)
The results are collated, drawing on theory to form a logical and coherent
215
What are the 6 main discursive strategies?
Nomination
Predication
Argumentation
Perspectivisation
Intensification
216
What are the two main approaches we can split critical discourse analysis into?
Three-dimensional approach (focus on the three stages of describing the text THEN undergoing the discursive practice of producing and consuming text THEN explaining the social context)
Discourse historical approach (identification of contents/tropics THEN the investigation of discursive strategies (e.g. argumentation) THEN the exploration of textual features and techniques on a level of micro-level analysis)
Not really competing approaches BUT different aspects of discourse
217
What are the 6 stages of critical discourse analysis?
1. Identify a discourse related to an injustice or inequality in society
2. Locate and prepare data sources
3. Explore the background of each text
4. Code text and identify over-arching themes
5. Analyse both the external and internal relations in the text
6. Interpret the meanings
218
What are the critiques of CDS/CDA? What are the corresponding defences?
Ideological commitment/lack of objectivity!... BUT Ideological commitment is part of the explicit agenda and does not equal analytical distortion
Too much methodological diversity... BUT methodological pluralism can be seen as a strength rather than a weakness
Interpretation rather than analysis... BUT the kind of interpretive work that CDS/CDA offers is closer to explanation than subjective understanding
219
What is quantitative text analysis?
Converting text into numerical values and the use of statistical analysis to identify patterns, trends and relationships within the text
220
What is computational text analysis?
Using AUTOMATED and semi-automated computational techniques to process, analyse and interpret textual data
221
What would be an example of hybrid quantitative text analysis?
Dictionary analysis
222
What are 2 examples of purely quantitative text analysis?
Statistical summary/analysis
Machine learning (computational analysis)
223
What are word clouds? (descriptive quant text analysis)
Collection of words displayed with a the words scaled to size based on frequency - can be divided into true and false poles
224
What are some examples of how we can use descriptive statistics for quant text analysis?
-Absolute word/feature frequency
-Relative word/feature frequency
225
What is the 'Key Words in Context' strategy in descriptive quant text analysis?
You gather a list of keywords, identifying the source and text and the word index number within the source text - then identify the context in which the key words appear
(can be sued to identify paragraphs of interest)
226
What is keyness? (descriptive quant text analysis)
You compare the different associations that keywords have i.e. with a target/reference group or actor
Helps you to understand which words are used more by one group (or about one group) relative to the other one
(e.g. can compare how often words are used by Obama vs Trump - Obama uses 'us' much more than Trump who uses 'America' a lot more)
227
What is a lexical dispersion plot? (descriptive quant text analysis)
A chart on which the occurrence of particular terms throughout texts are visualised - not only how often the term is used, but also where in the speech it is used
228
What is lexical diversity?(descriptive quant text analysis)
A measure of how many different words are used in a text - i.e. a measure of the richness of a text's vocabulary
229
What is lexical density? (descriptive quant text analysis)
A measure of the proportion of lexical items (i.e. nouns vs verbs vs adjectives vs adverbs) in a text - i.e. a measure of how complex the text is
230
What is co-occurrence? (descriptive quant text analysis)
A measure of co-occurrence (occurrence together) of features within a defined context - e.g. a document or window within a collection of documents
Helps us understand semantic fields and interconnected terms across literature - i.e. does 'islam' often co-occur with 'immigration' or 'terrorism' in political speeches?
Can be plotted as a co-occurrence network
231
When do we use descriptive statistics for text?
-When trying to present the characteristics of a corpus or collection within it
-For exploratory analysis (understand text...)
-Comparing texts - different texts, texts over time, between authors...
-When trying to measure frequency of concepts within text
Don't actually test causal relations! We can evaluate observational relations between variables and get close to capturing causal relations
232
Is a dictionary in quant text analysis actually a thesaurus?
Yes
233
Is dictionary analysis more than just counting?
Yes - you predefine words associated with specific meanings - assign them a category - THEN you find the absolute or relative rate at which key words appear in a text
234
What are the three things that we use dictionary analysis for?
1 - Measuring the prevalence of a concept in a text
2 - Measuring the extent to which documents belong to certain categories
3 - Classify documents into categories (based on the prevalence of categories)
235
What is the linguistic inquiry and word count dictionary analysis method?
It uses a dictionary to calculate the percentage of words in the text that match each of up to 82 language dimensions - e.g. positive or negative emotion, tense of verb...
236
What is the Harvard General Inquirer dictionary analysis method?
2 large sentiment categories - positive and negative
237
What 3 things do you have to think about when building a useful dictionary? (V-S-S)
Validity - is the category scheme valid?
Sensitivity - does the dictionary identify all relevant content?
Specificity - does it identify only the content that belongs there?
238
What is the hierarchy of levels in a dictionary analysis coding scheme?
First level - main domain (e.g. overarching idea of populism)
Second level - subdomain (e.g. idea of 'corrupt elites' as a theme within populism)
May be other further levels and additional sub-domains
239
As a general idea, what do we use dictionary analysis in quant text analysis to do?
Describe text - and within that identify words within categories and use identified features to compare texts
240
What are some advantages of dictionary analysis within quant text analysis?
Easy to use (computationally simple) AND easy to present and for readers to understand
Time efficient
Cheap
Powerful
241
What are some disadvantages of dictionary analysis within quant text analysis?
Domain specific - only good if the scores attached to words are aligned with the context in which the words are used
Time specific - the meaning of words can change over time
Only really good for text that uses standardised language - not great for text that uses sarcasm, irony, double meanings, and poor with figures of speech - all of these can mess up analysis
242
What is the idea of supervised classification in computational text analysis methods?
You as the researcher code a subset of the data - then use a supervised classifier to learn the relation between the words and the labels/categories - then the computer infers the labels for the rest of the dataset
You set the precedent for the computed text analysis...
243
What are the steps of supervised classification (comp text analysis)?
Data collection and labelling/coding
Data preprocessing (clean and prepare data for analysis)
Feature extraction (transform words into numbers)
Data splitting (separate human coded/labelled data into training and testing set)
Model selection, training and evaluation
Model deployment
Monitoring and maintenance
244
What is unsupervised classification in computational text analysis?
You discover the main themes/topics in an unstructured corpus - organise the collection according to said themes
Requires no prior info, training... requires decision about the number of topics before though
245
What is the structured topic model
It shows how some covarities (e.g. the liberalism or conservatism of somebody) are associated with the prevalence of topic usage (like talking points almost)
The relationship between covariates and latent topics is most frequently the quantity of interest
246
Which method (descriptive methods, dictionary analysis and computational text analysis) is best in quantitative text analysis?
The best method can only be determined in relation to a specific research project - depends on a range of factors and especially your research question which collection and analysis methods should be tailored to
Can combine methods if best for RQ!
247
What is secondary data?
Any data that is collected by others (i.e. not yourself) at an earlier point in time
(primary data is data that we collect ourselves; original information collected by the researcher themselves)
248
What are expert coded datasets?
Datasets where experts have already provided estimates and assessments of various measures; usually opinion based
249
What are researcher coded datasets?
Datasets where teams of researchers have already coded them using publicly available information (i.e. news sources, academic articles, etc...) - like UCDP where researchers code data on conflicts and casualties
250
Can surveys be used for secondary data research?
Yes - there are many pre-existing high quality academic surveys - e.g. European social survey which measures voter opinions across countries
251
Can official statistics be used for secondary data research?
Yes - they are conducted by national governments, IGOs, and independent organisations (e.g. electoral commission)
252
Can quantified text be used for secondary data research?
Yes - codifiable text e.g. party manifestos, tweets, parliamentary speeches, policy documents, news stories, etc...
253
Can qualitative sources be used for secondary data research?
Yes - any existing sources of qualitative data that can be analysed with qualitative methods (i.e. content analysis) can be used for secondary data research - e.g. government reports. newspapers...
254
Can data repositories be used for secondary data research?
Yes - we can also find published datasets used by researchers in their own work - researchers often upload versions of the datasets they use in published papers for transparency, reproductibility, and replicability
255
What are 3 main things to look for in exploring secondary datasets?
VARIABLES - columns in our dataset that measure certain characteristics (e.g. gender, age...)
CODEBOOK - guide that lists all variable names, what they measure, and the range the variable can tale on
FILE FORMATS - which file extension does the dataset use? Important for importing into different statistical software programs (e.g. stata, R, python...)
256
What are continuous variables?
A variable where entities get a distinct score... two main types...
Interval variable: Equal intervals on the variable represent equal differences in the property being measured (i.e. the difference between 6 and 8 is equivalent to the difference between 13 and 15) - each interval is equal
Ratio variable: The same as an interval variable, but the ratios of scores on the scale must also make sense (i.e. a score of 16 on an anxiety scale means that the person is actually twice as anxious as someone scoring 8).
257
What are 4 benefits of using secondary data?
Quality - secondary data is often of higher quality (done by experts or professional researchers), larger scale... often not feasible to collect their own data
Longer time scale - because lots of data collection efforts are part of larger projects, they are often collected over periods of time as opposed to single surveys that capture a moment in time
Cost efficient - collecting data is costly
Valid and reliable - often have a pre-established degree of validity and reliability - maybe wet through peer-review
258
What are 3 disadvantages of using secondary data?
Unable to make causal inferences - because secondly data sources are often not experimental and so it is difficult to make causal inferences
Inaccuracies - if there are inaccuracies in the data and how it was coded or collated, these will be replicated in your research if you use the data
Limitations on samples - samples and populations covered may be limited and may not cover the populations you wish to research (may be limitations on what data is collected - i.e. may not include the variables you are interested in)
