What are the problems with selecting on the dependant variable? Have thee problems been overstated?
Selecting on the dependant variable, not enough attention is payed to the independent variable and important factors are excluded from the analysis.
Overstated: been labelled an ‘inferential felony’ to do so
If the cases you select affect the answers you get, does this mean that small N comparison is inherently less reliable than large N comparison.
Small N: better at generating theory than testing due to selection bias being a likley problem - makes it hard to make robust generalizations
A hypothesis must be tested across a wide number of cases or else findings have to be treated with caution.
What is Most similar systems design? (MSSD)
Based on selecting countries that share important (theoretical) characteristics but differ on one crucial aspect (related to hypothesis)
Frequently used in area studies
Shared characteristics act as a control in order to test whether the crucial difference is associated with the variation in the dependant variable.
What is most different systems design? (MDSD)
Selecting cases that are different in most respects and only similar on the key explanatory variable of interest.
In this case the differences act as a control.
Selections are usually based on the independent variable and not on the dependent - Some probs associated with this.
What is concept stretching? To what extent can it hinder comparative research?
- Defining concepts in very broad terms in order to ensure that competing definitions or interpretations of the concept overlap with each other.
- Conceptual definitions and measures can mean different things to people and countries so careful analysis and substantiative knowledge is key in comparative research.
How does QCA differ from small N research?
- QCA occupies and intermediate place between SN and LN (12-100) and uses quantitative methods to examine causal complexities.
Small N uses systemic qualitative analysis of certain countries or groups.
What are the 2 ways of doing QCA?
- Crisp set QCA: all variables in the analysis are treated as simple dichotomies - it either is or isn’t the thing
- Fuzzy set QCA: All the variables are allowed to take different values and are calibrated on an interval scale between 0.0 and 1.0 - The amount that the variable is included is given a numerical value.
What is QCA - general characteristics.
- Heuristic tool - summarising data, producing typologies and elaborating new theories or models.
- Well suited to unravelling causal complexity in order to detect the different conditions that can lead to the same outcome.
Boolean algebra
- Used for examining all the possible combos of variables that are associated with an outcome and trying to simplify these to a as few necessary combos as possible
QA with large N - points to note.
- Allows for rigorous testing on diff hypotheses and make inferences about how variables are connected.
- Allows for examining how different factors interact with each other and produce different consequences in different contexts.
- Generally relies on quantitative data
- Surveys, cross-national data sets, policy docs, speeches
Geddes limitations of SN and LN
- Analysts trying to explain developing countries growing faster economically than others focus on new industrialising countries.
- Noted that a few countries had repressed labour forces and it was asserted that this constituted a comparative advantage in international economic comp.
- Few cases and sweeping generalisations - good for theory(Small -n)
- To test Geddes collected data on economic growth and labour repression for 84 dev countries.
Equivalence of meaning.
Whether the theoretical concepts and empirical indicators of those conceps mean the same things and measure the same things in the different contexts.
- e.g populism - can mean different things in different places - we think we are comparing the same thing but nope.
