Research logic and research design Flashcards
Two types of research
- Normative
- Positive
The two inform themselves, you can do positive research and then do normative research to find what works better.
Normative research
How the world should be, value judgements, ethics, requires application of philosophy rather than data. The aim is prescriptive.
Positive research
What is/was/will be, deals with empirical phenomena and theoretical concepts and the links between them, avoiding biased judgements, capable of research through data collection. Aim is to avoid biased judgements.
Two types of positive research
- Theoretical: Theory elaboration, conceptualisations, using logical deductive reasoning to develop truth statements (analytical).
- Empirical: Theory generating, testing/application and conceptualisation. Truth statements result from a confrontation between theory and empirical content (synthetic).
Three types of empirical research
- Descriptive
- Predictive
- Explanatory
Descriptive research
Describing what’s going on. Collection of relevant facts that can be used as evidence in theory building or concept formation. Can lead to a “light bulb” moment. This light bulb moment (patterns) can inform explanatory research. Case centred.
Predictive research
Predictions about the future through identifying future patterns. Oriented towards elections. Not focused on in the course.
Explanatory research
Explaining why or how is it going on. Can focus on causes of events, causal effects or causal mechanisms.
- Why questions about causal effects
- How questions about causal mechanisms
Explanatory and descriptive research are not exclusionary, they are often combined within one study at different stages.
Can inform descriptive research when more information is needed.
What is research design?
- A logical plan for getting from the research question to the answer.
- Not just about how you do the research, but also why and what the purpose of doing this is.
- Answers the question of how are you going to conduct the research
How to choose a good research design?
- Make decisions based on the theory out there
- In explanatory research, you have to think about the variation you want to explain
- The level of analysis
- The type of data you have to collect to answer your research question
- If there’s a probabilistic or deterministic causational perspective
- Think about the choice of methods
Theory testing
When there is research you can build on
1. Start with describing/analysing the theory.
2. Hypotheses/propositions
3. Measurement/sampling etc.
4. Data collection
5. Data analysis
6. Either the data analysis makes it so there’s implications for the hypothesis, so you confirm/reject the theory. When you reject the theory, a new theory is needed.
Deductive (when theory testing and theory building, abductive)
Theory building
When there is not much or no theory to build on.
1. Start with data collection
2. Data analysis
3. Implications for hypothesis or new theory
4. When implications for hypothesis, it results in a theory
5. When new theory is needed, a new hypothesis is needed and then you go through the process of measurement, data collection, data analysis and implications for hypothesis again and making a hypothesis in the end.
Inductive (when theory testing and theory building, abductive)
Variation
Which variation in X or Y do we want to explain?
1. In empirical explanatory research
2. The variation we are interested in are patterns, similarties and differences.
3. Explanations can focus on causal factors, causal effects or causal mechanisms.
Types of variation
- X-centred studies focus on a cause and asks whether it has an effect on a given outcome (causal factors). It looks for the contribution of X in explaining part of the variation in Y.
- Y-centred studies focus on the outcome and seeks to find out the relevant causes (causal effect). Explaining the variation in Y as best as you can.
- Mechanism-centred studies focus on tracing a causal mechanism or a causal process. Uncovering the sequence of intervening factors that link an X to a Y.
What is a case?
- A unit of analysis in which we measure variation.
- Representation of some sort of population.
- Empirical phenomenon that is an example of a population.
- It has boundaries, is spatial, temporal and substantive, based on this we can tell what the population is.
- When generalising you’re looking for causal homogeneity, meaning causal effects and mechanisms are expected to hold true for other cases in the population
What is a case study?
An empirical analysis of a small sample.
Sample is about studied cases, population is studied and unstudied cases (what you want to generalise).
What are units of variation?
In explanatory research.
Small N designs or large N designs, based on the amount of cases.
Types of research designs
- Single case, short time and short space (one case in time)
- Comparative case study (different case studies at a specific time point)
- Periodisation study (one case across different points in time)
Analytical level (level of analysis)
On which analytical level do we observe our research targets?
1. Macro-level (societies, economies, states, etc.)
2. Meso-level (groups, territorial subunits, etc.)
3. Micro-level (individuals)
Level of analysis
On which level of analysis does the argument take place.
1. Cross-case level (causal effects). All quantiative research because it is large N.
2. Within-case (causal mechanisms).
Coleman’s bathtub
- Explanatory factor (X) —> Outcome (Y).
There is an argument for how X leads to Y: Explanatory factor X influences actors who behave in a certain way, which leads to the outcome.
The explanatory factor and outcome are on macro-level or meso-level, the actors and behaviour on micro-level. - In causal effects and with more than one case, the X and Y (transformational mechanism) will be on cross-case level (situational mechanism). To understand the link between X and Y, you look at the within-case level (action-formation mechanism).
Two perspectives towards causation
- Probabilistic
- Deterministic
Probabilistic perspective towards causation
When the values of an independent variable increase or decrease, this usually results in the values of the dependent variable increasing or decreasing
Cause is a probability raiser
Causation
- A type of co-variation where one phenomenon contributes to or produces another.
- Causal relationships are the core of explanatory research.
- You cannot observe causation, you can work out causation from empirical observations
Criteria for causal inference based on empirical observations
- Temporal sequence (X –> Y)
- Proximity (in time and space)
- Constant coexistece of X and Y
- Logical connection between X and Y
You also need to rule out other factors that could influence the outcome.
Deterministic perspective towards causation
- When the values of an independent variable increase or decrease, this always results in the values of the dependent variable increasing or decreasing
- Explanatory factors are necessary and/or sufficient conditions for an outcome
Necessary condition in deterministic perspective towards causation
Something which must be present for something else to be possible.
Example: it is necessary to run in an election as a candidate to be able to win a seat in parliament
Sufficient condition in deterministic perspective towards causation
Something which, if present, guarantees that something else will occur.
Example: holding new elections is a sufficient condition for government determination
How to evaluate a research design
- Positivist research designs should have internal and external validity. Internal meaning the structure of the design makes it to draw unambiguous conclusions from the results and external meaning that it can be generalised.
- Interpretivist research designs should be credible, meaning the analysis is authentic and offers a believable interpretation of reality. It should also be transferable, meaning readers can assess the broader applicability of the lessons drawn from the findings.
