Lecture 1
Research approaches:
Deductive approach (top-down): Theory Develop Hypotheses Gather data Feedback to
theory
- Used in confirmatory research
- Relies more on quantitative studies
- Test theories
- Structured and systematic
Inductive approach (bottom-up): Start with data Develop hypothesis
Form Theory
- Used in exploratory studies
- Relies more on qualitative studies
- Develop theories
- Flexible to new findings
Research aims:
Exploration: unknown topic, familiarisation, initial understanding (e.g.what is the
phenomenon made up of?)
Description: precisely measure and describe characteristics of variables/phenomena (e.g. To
what extent is the phenomenon present here?)
Explanation: studying causes for phenomena, relationships, incl. causal ones (e.g. Why is
the
phenomenon most prevalent in this case?)
Evaluation: evaluates a program, project, intervention (e.g. To what extent was this project
effective in countering misinformation?)
Studying (international) crime data, challenges:
Larger dark number: many crimes not recorded, shame and stigma for reporting.
- Double dark number (if authorities are the perpetrators or involved hide the numbers)
No sampling frame for offenders or victims (we don’t know who has or will offend or be
victimized)
Potentially highly politicized setting
Ethical and cultural issues in working with offenders and victims
Due to the large-scale nature of events, reliability of data might be questionable
Archives may not be available in conflict or post-conflict settings
Researcher may have no access to the field
Conceptualization and operationalization aspects increase the reliability and validity of
the study:
Conceptualization: specify the meaning of the concepts and variables to be studied
aim: to produce a specific agreed-on meaning for a concept
e.g. religiosity: when is someone religious? make use of indicators: an observation that acts as
a reflection of a variable we wish to study e.g. attending religious services as an indicator of
religiosity.
Operationalization: how will we actually measure the variables under study?
Dimension: groups of indicators under common themes e.g. “religiosity” might be specified in
terms of a belief dimension, a ritual dimension, a devotional dimension, a knowledge dimension,
and so forth.
Lecture 2
Types of variables:
Nominal: categories, little information (e.g. A,B,C)
Ordinal: ranked categories, but ranked order (e.g. A>B>C)
Interval: equal distances between intervals, can compare them
(e.g. temperature, IQ) doesn’t have a ‘true’ zero
Ratio: there is a ‘true’ zero possible to compare ratios (e.g.
weight, height)
Mixed Method Designs - sequential or parallel
Sequential Design: e.g. qualitative study first develop
hypotheses test
quantitatively [confirmation; explanation; help generalize
findings]
, Parallel Design: qualitative and quantitative simultaneous integrated findings
[complementarity, compare or relate]
Each directly affects the quality of inferences drawn from the entire mixed methods study
.
Reliability (or dependability): accuracy of the measurement method consistently the same
results (over time).
Extent to which the research process is logical, traceable, and clearly documented and that
the methods chosen and the decisions made by the researchers are transparent.
Reliability of quantitative studies:
Test-retest method: Consistency of measure over time (e.g. reconducting the same survey
later)
Using established measures: Increases the chances that measure can be assessed with
test-retest and split-half methods (results over time and across studies can be compared)
Multiple raters (inter-rater reliability): Consistency of a measure across raters or observers,
have two individuals code the same text and calculate the degree of agreement (improves
precision)
Validity or Credibility ( “trustworthiness”): Whether the findings hold true
Internal validity (Truth in Study): are we measuring what we intended to measure?
- In Quantitative: degree of confidence in the findings of a study (not influenced by
confounding factors?)
- In Qualitative: the extent to which the representation of data fits the views of the
participants
External validity/ transferability (Truth in Real life): the extent to which results from a study
can be
applied to other contexts - situations, groups or events - (is it generalizable?) .
How well does a measurement reflect the true meaning of the concept being studied?
First step Face validity (prima facie validity): does the measure make sense? Is it
reasonable?
Further Construct validity (aspects):
- Content validity: Does the measure cover all aspects of the concept?
- Criterion validity: Is the measurement connected to or can it predict a real-world outcome
(e.g. if a test is designed to measure job performance, it should correlate with how well
someone actually performs on the job)
- Construct validity in-the-narrow-sence: Is the measure designed fairly, without bias
against any group of participants?
1st Reliable due to its consistency, but not valid because
way of the centre
2 Valid because some are towards the centre, but not
realibe as it is not consistent
3 Both
Accuracy: whether the response is a correct reflection
of the real world
Increasing precision might reduce your accuracy (e.g. asking to date an event by exact day
is more precise, but if it is likely that the individuals might have forgotten (e.g. long time
ago, not crucial event, or in a culture where calendar dates are less used), it will give you
less accurate responses.
However, improving precision improves the reliability of the methods.
Reflexivity: Role of the researcher is made clear (Reflecting on own bias, effect of non-
respondents, etc)
Confirmability: how well the findings of a study are supported by the actual data collected,
ensuring that the results truly reflect the participants' views rather than the researcher’s personal
biases or assumptions (neutrality, objectivity).
To enhance confirmability, include direct evidence from the data (e.g. quotations)
Consent form: an agreement made between the researcher and the respondent.
Seminar 1:
Common research methods in criminology:
Obtrusive research (involving participants): Interviews/survey research/Focus group -
research on attitudes, characteristics, opinions, experience) - Participatory research -
, research on group structures, patterns, exploration of a phenomenon, situation where the
context is important)
Unobtrusive research: content analysis; analysis of secondary data; literature reviews,
meta-analysis
Questions structure:
Structured: pre-determined, more reliable, no spontaneous follow-up, suitable for larger
projects,
- Benefit: reduces the influence of the interviewer
Semi-structured: Topic-guided, but flexible, room for follow-up questions, useful when data
collection has clear aims and limited time
- Benefit: less exhausting for the respondents
Unstructured: conversation-like, in-depth, timely, requires good interview skills, and lower
reliability
- Benefit: high internal validity,
Issue of reactivity: subjects of research may react to fact of being studies and alter their
behaviour
Lecture 3
Sampling aims:
Quantitative studies: aiming at representativeness & generalizability
Qualitative studies: aiming at maximizing information, until saturation: additional
respondents do not add new relevant information (so representativeness to the overall
population not an issue)
Thus, different aims and therefore different strategies
Important whichever collection method is chosen: Determining the population and sampling
approach
Types of samples:
Probability samples: representative
- Random, Systematic, Stratified, and Cluster, sample
Good sample reflects the charactersitics of the population (is representative) External
validity
Sampling frame: list of all population members
Complete sampling frame + random sample of sufficient size would be ideal!
When are samples random (or probabilistic)? Litmus test: when every population member
has an equal chance to end up in the sample Otherwise: under- or overrepresentation of
certain groups
Simple random sample: sample members are drawn randomly from a complete pre-existing
sampling frame
Can be inefficient and costly
No control over spread and distribution
Systematic sample: Random starting point with fixed intervals
Is efficient and ensures spread, but not distribution
Randoms when the starting point is random and not the same periodicity (e.g. only
corner houses)
Stratified sample: The population is divided into groups (strata, e.g. male/female) and
random sample members are taken from each stratum
Ensures representation of each strata and good distribution
Reweighting could be used is sample is disproportionate
Cluster sample: The population is divided into smaller groups =clusters (e.g. based on
geographical area), and entire clusters are randomly selected.
Very efficient, can be used for large populations
But amplifies sampling error:
- Clusters tend to have similar individuals (e.g., people in the same neighborhood often
share similar traits).
As a result, the data can be "noisier," meaning less precise and less representative of the entire
population.
Research approaches:
Deductive approach (top-down): Theory Develop Hypotheses Gather data Feedback to
theory
- Used in confirmatory research
- Relies more on quantitative studies
- Test theories
- Structured and systematic
Inductive approach (bottom-up): Start with data Develop hypothesis
Form Theory
- Used in exploratory studies
- Relies more on qualitative studies
- Develop theories
- Flexible to new findings
Research aims:
Exploration: unknown topic, familiarisation, initial understanding (e.g.what is the
phenomenon made up of?)
Description: precisely measure and describe characteristics of variables/phenomena (e.g. To
what extent is the phenomenon present here?)
Explanation: studying causes for phenomena, relationships, incl. causal ones (e.g. Why is
the
phenomenon most prevalent in this case?)
Evaluation: evaluates a program, project, intervention (e.g. To what extent was this project
effective in countering misinformation?)
Studying (international) crime data, challenges:
Larger dark number: many crimes not recorded, shame and stigma for reporting.
- Double dark number (if authorities are the perpetrators or involved hide the numbers)
No sampling frame for offenders or victims (we don’t know who has or will offend or be
victimized)
Potentially highly politicized setting
Ethical and cultural issues in working with offenders and victims
Due to the large-scale nature of events, reliability of data might be questionable
Archives may not be available in conflict or post-conflict settings
Researcher may have no access to the field
Conceptualization and operationalization aspects increase the reliability and validity of
the study:
Conceptualization: specify the meaning of the concepts and variables to be studied
aim: to produce a specific agreed-on meaning for a concept
e.g. religiosity: when is someone religious? make use of indicators: an observation that acts as
a reflection of a variable we wish to study e.g. attending religious services as an indicator of
religiosity.
Operationalization: how will we actually measure the variables under study?
Dimension: groups of indicators under common themes e.g. “religiosity” might be specified in
terms of a belief dimension, a ritual dimension, a devotional dimension, a knowledge dimension,
and so forth.
Lecture 2
Types of variables:
Nominal: categories, little information (e.g. A,B,C)
Ordinal: ranked categories, but ranked order (e.g. A>B>C)
Interval: equal distances between intervals, can compare them
(e.g. temperature, IQ) doesn’t have a ‘true’ zero
Ratio: there is a ‘true’ zero possible to compare ratios (e.g.
weight, height)
Mixed Method Designs - sequential or parallel
Sequential Design: e.g. qualitative study first develop
hypotheses test
quantitatively [confirmation; explanation; help generalize
findings]
, Parallel Design: qualitative and quantitative simultaneous integrated findings
[complementarity, compare or relate]
Each directly affects the quality of inferences drawn from the entire mixed methods study
.
Reliability (or dependability): accuracy of the measurement method consistently the same
results (over time).
Extent to which the research process is logical, traceable, and clearly documented and that
the methods chosen and the decisions made by the researchers are transparent.
Reliability of quantitative studies:
Test-retest method: Consistency of measure over time (e.g. reconducting the same survey
later)
Using established measures: Increases the chances that measure can be assessed with
test-retest and split-half methods (results over time and across studies can be compared)
Multiple raters (inter-rater reliability): Consistency of a measure across raters or observers,
have two individuals code the same text and calculate the degree of agreement (improves
precision)
Validity or Credibility ( “trustworthiness”): Whether the findings hold true
Internal validity (Truth in Study): are we measuring what we intended to measure?
- In Quantitative: degree of confidence in the findings of a study (not influenced by
confounding factors?)
- In Qualitative: the extent to which the representation of data fits the views of the
participants
External validity/ transferability (Truth in Real life): the extent to which results from a study
can be
applied to other contexts - situations, groups or events - (is it generalizable?) .
How well does a measurement reflect the true meaning of the concept being studied?
First step Face validity (prima facie validity): does the measure make sense? Is it
reasonable?
Further Construct validity (aspects):
- Content validity: Does the measure cover all aspects of the concept?
- Criterion validity: Is the measurement connected to or can it predict a real-world outcome
(e.g. if a test is designed to measure job performance, it should correlate with how well
someone actually performs on the job)
- Construct validity in-the-narrow-sence: Is the measure designed fairly, without bias
against any group of participants?
1st Reliable due to its consistency, but not valid because
way of the centre
2 Valid because some are towards the centre, but not
realibe as it is not consistent
3 Both
Accuracy: whether the response is a correct reflection
of the real world
Increasing precision might reduce your accuracy (e.g. asking to date an event by exact day
is more precise, but if it is likely that the individuals might have forgotten (e.g. long time
ago, not crucial event, or in a culture where calendar dates are less used), it will give you
less accurate responses.
However, improving precision improves the reliability of the methods.
Reflexivity: Role of the researcher is made clear (Reflecting on own bias, effect of non-
respondents, etc)
Confirmability: how well the findings of a study are supported by the actual data collected,
ensuring that the results truly reflect the participants' views rather than the researcher’s personal
biases or assumptions (neutrality, objectivity).
To enhance confirmability, include direct evidence from the data (e.g. quotations)
Consent form: an agreement made between the researcher and the respondent.
Seminar 1:
Common research methods in criminology:
Obtrusive research (involving participants): Interviews/survey research/Focus group -
research on attitudes, characteristics, opinions, experience) - Participatory research -
, research on group structures, patterns, exploration of a phenomenon, situation where the
context is important)
Unobtrusive research: content analysis; analysis of secondary data; literature reviews,
meta-analysis
Questions structure:
Structured: pre-determined, more reliable, no spontaneous follow-up, suitable for larger
projects,
- Benefit: reduces the influence of the interviewer
Semi-structured: Topic-guided, but flexible, room for follow-up questions, useful when data
collection has clear aims and limited time
- Benefit: less exhausting for the respondents
Unstructured: conversation-like, in-depth, timely, requires good interview skills, and lower
reliability
- Benefit: high internal validity,
Issue of reactivity: subjects of research may react to fact of being studies and alter their
behaviour
Lecture 3
Sampling aims:
Quantitative studies: aiming at representativeness & generalizability
Qualitative studies: aiming at maximizing information, until saturation: additional
respondents do not add new relevant information (so representativeness to the overall
population not an issue)
Thus, different aims and therefore different strategies
Important whichever collection method is chosen: Determining the population and sampling
approach
Types of samples:
Probability samples: representative
- Random, Systematic, Stratified, and Cluster, sample
Good sample reflects the charactersitics of the population (is representative) External
validity
Sampling frame: list of all population members
Complete sampling frame + random sample of sufficient size would be ideal!
When are samples random (or probabilistic)? Litmus test: when every population member
has an equal chance to end up in the sample Otherwise: under- or overrepresentation of
certain groups
Simple random sample: sample members are drawn randomly from a complete pre-existing
sampling frame
Can be inefficient and costly
No control over spread and distribution
Systematic sample: Random starting point with fixed intervals
Is efficient and ensures spread, but not distribution
Randoms when the starting point is random and not the same periodicity (e.g. only
corner houses)
Stratified sample: The population is divided into groups (strata, e.g. male/female) and
random sample members are taken from each stratum
Ensures representation of each strata and good distribution
Reweighting could be used is sample is disproportionate
Cluster sample: The population is divided into smaller groups =clusters (e.g. based on
geographical area), and entire clusters are randomly selected.
Very efficient, can be used for large populations
But amplifies sampling error:
- Clusters tend to have similar individuals (e.g., people in the same neighborhood often
share similar traits).
As a result, the data can be "noisier," meaning less precise and less representative of the entire
population.
