Session 5
I – Deductive reasoning
Deductive-nomological explanation : the phenomenon to be explained (explanandum) is
logically deducible from that which does the explaining (the explanans); and it is nomological because
the explanans includes at least one law (‘nomos’, Greek).
According to this model, something is explained when it is shown to be a member of a more
general class of things, when it is deduced from a general law or set of laws.
Theory serves as a point of departure in research offers relevant concepts that are defined in the
context of the research and used as a basis for explaining observed phenomena.
Deduction : ‘from the general to the particular/specific’
Characteristics Usually in positivist research
Interested in cause-effect relationships
Research designed to fend off alternative explanations (falsification)
Approach 1- From theory to concept(s)
2- Hypothesis (proposition to predict a relationship between variables, if A then B)
3- Data collection
4- Data analysis
5- Rejection/confirmation of hypothesis
When testing cause-and-effect relationships, validity can be internal or external.
Are the measures suitable as indicator of the central concept(s)?
Face validity : Does the indicator make sense intuitively? Is there a broad agreement that the indicator
is directly relevant to the concept?
Content validity : Does the indicator cover all the dimensions of a concept? It requires precision in the
concept definition (or to restrict the scope of the study).
Construct validity : Does the measure conform to our theoretical expectations in terms of its association
with other relevant factors?
Convergent validity (sub-type of construct validity) : Does the indicator correlate with measurements of
the same variable if measured with a different instrument? Established by correlation: Do different
questions that target the same construct correlate with each other?
Measurement reliability : Does repeated measurement with the same instrument lead to the same
results? Assuming that we measure the same thing under the same conditions and the first measurement
does not affect the results of the second.
Inter-observer reliability Are different observers consistent with the estimates of the phenomenon?
Test-retest reliability Are measures consistent if done later in the same way?
We accept that we will always have some measurement error.
> Random error (unreliable)
> Systematic error (bias)
A – Experiment
Experiments are meant to model situations after natural sciences : controlled circumstances lead
to controlled interventions because it is possible to allow for changes only in the variable on which the
focus is. Only a limited number of social phenomena can be investigated via experiments because of a
, too vast potentiality of variables that may influence any phenomenon, but also for ethical
considerations.
B – Observational designs
“The statistical method”, its is less reliable and valid than an experiment. Still, it provides good
tests for eliminating rival theories. There is a need for large samples to have a statistically significant
sample under study, however, the number of variables that could ‘disturb’ the statistics is higher.
Population All the units (‘cases’) that your claim is said to apply to
Sample A subset of the population that is investigated
N (sample size) The number of units in the sample
Cross-sectional design
Multiple observations (cases), large sample
On one particular moment in time
Comparing variation among different samples
Focus on dependent variable, the independent variable is its explanation
Causality cannot be proven, observations may reveal correlations, but it is different from causation
Longitudinal design
Multiple observations (cases), large sample
On several moments in time
Using one and the same sample (“attrition”)
> Cohort study, focusing on a group of people with specific traits which are constant over time
> Panel study, focusing on a random sample and collecting information on two or more points in time
Repeated cross-sectional studies have a ‘longitudinal character’, it may indicate societal trends
C – Comparative case research
It is an alternative to using statistics, if the number of cases is too low for statistical inferences
it is still possible to establish general/empirical relations between two variables. Comparison is made
for the most important independent variables, by keeping all other variables as constant as possible.
Comparative case research adopts the same logic as statistical research yet dealing with many variables
and a smaller sample.
Generalizability (external validity) or transferability helps with drawing on existing theory and
concepts, validity and reliability depend on following a set of rules of scientific inference to fend off
alternative explanations with conviction. Sometimes it is not possible to know what is unknown and
some concepts could be made more or less “sensitizing”.
II – Inductive reasoning
Knowledge production via induction is from the particular to the generic’. In general, common
sense ideas serve as a point of departure. Data collection and analysis help generate concepts. Research
process is iterative: “weaving back and forth between data and theory”. Reflection leads to concept
formation and to the formulation of a theory, which is the resultant of research.
A fully inductive approach, or ‘grounded theory’ has the general following from the particular.
However, “one never enters the field intellectually empty-handed” since ideas are the point of
departure: with ‘sensitizing concepts’ (whereas definitive concepts provide prescriptions of what to
see, sensitizing concepts merely suggest directions along which to look).
I – Deductive reasoning
Deductive-nomological explanation : the phenomenon to be explained (explanandum) is
logically deducible from that which does the explaining (the explanans); and it is nomological because
the explanans includes at least one law (‘nomos’, Greek).
According to this model, something is explained when it is shown to be a member of a more
general class of things, when it is deduced from a general law or set of laws.
Theory serves as a point of departure in research offers relevant concepts that are defined in the
context of the research and used as a basis for explaining observed phenomena.
Deduction : ‘from the general to the particular/specific’
Characteristics Usually in positivist research
Interested in cause-effect relationships
Research designed to fend off alternative explanations (falsification)
Approach 1- From theory to concept(s)
2- Hypothesis (proposition to predict a relationship between variables, if A then B)
3- Data collection
4- Data analysis
5- Rejection/confirmation of hypothesis
When testing cause-and-effect relationships, validity can be internal or external.
Are the measures suitable as indicator of the central concept(s)?
Face validity : Does the indicator make sense intuitively? Is there a broad agreement that the indicator
is directly relevant to the concept?
Content validity : Does the indicator cover all the dimensions of a concept? It requires precision in the
concept definition (or to restrict the scope of the study).
Construct validity : Does the measure conform to our theoretical expectations in terms of its association
with other relevant factors?
Convergent validity (sub-type of construct validity) : Does the indicator correlate with measurements of
the same variable if measured with a different instrument? Established by correlation: Do different
questions that target the same construct correlate with each other?
Measurement reliability : Does repeated measurement with the same instrument lead to the same
results? Assuming that we measure the same thing under the same conditions and the first measurement
does not affect the results of the second.
Inter-observer reliability Are different observers consistent with the estimates of the phenomenon?
Test-retest reliability Are measures consistent if done later in the same way?
We accept that we will always have some measurement error.
> Random error (unreliable)
> Systematic error (bias)
A – Experiment
Experiments are meant to model situations after natural sciences : controlled circumstances lead
to controlled interventions because it is possible to allow for changes only in the variable on which the
focus is. Only a limited number of social phenomena can be investigated via experiments because of a
, too vast potentiality of variables that may influence any phenomenon, but also for ethical
considerations.
B – Observational designs
“The statistical method”, its is less reliable and valid than an experiment. Still, it provides good
tests for eliminating rival theories. There is a need for large samples to have a statistically significant
sample under study, however, the number of variables that could ‘disturb’ the statistics is higher.
Population All the units (‘cases’) that your claim is said to apply to
Sample A subset of the population that is investigated
N (sample size) The number of units in the sample
Cross-sectional design
Multiple observations (cases), large sample
On one particular moment in time
Comparing variation among different samples
Focus on dependent variable, the independent variable is its explanation
Causality cannot be proven, observations may reveal correlations, but it is different from causation
Longitudinal design
Multiple observations (cases), large sample
On several moments in time
Using one and the same sample (“attrition”)
> Cohort study, focusing on a group of people with specific traits which are constant over time
> Panel study, focusing on a random sample and collecting information on two or more points in time
Repeated cross-sectional studies have a ‘longitudinal character’, it may indicate societal trends
C – Comparative case research
It is an alternative to using statistics, if the number of cases is too low for statistical inferences
it is still possible to establish general/empirical relations between two variables. Comparison is made
for the most important independent variables, by keeping all other variables as constant as possible.
Comparative case research adopts the same logic as statistical research yet dealing with many variables
and a smaller sample.
Generalizability (external validity) or transferability helps with drawing on existing theory and
concepts, validity and reliability depend on following a set of rules of scientific inference to fend off
alternative explanations with conviction. Sometimes it is not possible to know what is unknown and
some concepts could be made more or less “sensitizing”.
II – Inductive reasoning
Knowledge production via induction is from the particular to the generic’. In general, common
sense ideas serve as a point of departure. Data collection and analysis help generate concepts. Research
process is iterative: “weaving back and forth between data and theory”. Reflection leads to concept
formation and to the formulation of a theory, which is the resultant of research.
A fully inductive approach, or ‘grounded theory’ has the general following from the particular.
However, “one never enters the field intellectually empty-handed” since ideas are the point of
departure: with ‘sensitizing concepts’ (whereas definitive concepts provide prescriptions of what to
see, sensitizing concepts merely suggest directions along which to look).
