UNISA CMY3702 Criminological Research: Methodology and Data Analysis Notes

CMY3702, Criminological Research: Methodology and Data Analysis, equips students with the practical tools needed to plan, conduct, and analyze criminological research in ethically responsible ways. The module typically expects competence in research design choices, data collection strategies, measurement, sampling, and the interpretation of quantitative and qualitative evidence. These notes focus on methodology and data analysis with a strong criminology lens—covering how UNISA learners can develop credible research outputs, defend methodological decisions, and produce defensible findings.

1) Positioning CMY3702: What “Good” Criminological Research Looks Like

Criminological research is not just about “finding facts”; it is about producing trustworthy knowledge that can withstand academic scrutiny. In CMY3702, the emphasis is usually on whether the research design logically supports the research questions, whether the data are appropriately collected and measured, and whether the analysis methods match the nature of the evidence.

1.1 Core competencies assessed in criminological methodology

Across South African university assessments, criminology methodology typically tests the ability to:

• Translate a topic into research questions and objectives
  • Problem statement → research question(s) → objectives
• Choose a research design that fits the question
  • Quantitative, qualitative, mixed methods, or explanatory sequential designs
• Operationalize constructs
  • Turn abstract ideas (e.g., “perceived safety”) into measurable variables or themes
• Select ethical procedures
  • Participant protection, consent, confidentiality, risk minimization
• Sample appropriately
  • Probability sampling when generalization is needed; purposive sampling when depth/coverage is needed
• Collect data consistently and credibly
  • Instrument quality (validity/reliability), fieldwork management, documentation
• Analyze data with methodological coherence
  • Use appropriate descriptive, inferential, and qualitative analytical approaches
• Interpret responsibly
  • Distinguish correlation from causation, consider bias, and connect findings to theory and literature

CMY3702 is “methodology and data analysis,” so strong answers generally show methodological coherence: every methodological decision should connect logically to the purpose of the study.

1.2 Typical criminological research problem patterns

Many criminology students struggle because their methodology does not align with the problem pattern. Common problem patterns include:

1. Descriptive mapping
   • “What is the prevalence of street-level violence in area X?”
2. Explanatory modeling
   • “Do community disorder factors predict fear of crime?”
3. Process exploration
   • “How do offenders interpret rehabilitation interventions?”
4. Evaluation
   • “Did a diversion programme reduce reoffending?”
5. Comparative analysis
   • “How do perceptions of policing differ between youth and adults?”

Each pattern points to different methods. For example:

• Evaluation often needs quasi-experimental logic or strong comparison strategies.
• Process exploration is usually qualitative (interviews, observations, document analysis).
• Explanatory modeling is often quantitative (regression, logistic models).

1.3 Methodological coherence as a marking criterion

When examiners mark CMY3702-style answers, they often reward students who show coherence using a clear chain:

Research question → variables/themes → sampling → instrument/interview guide → data type → analysis technique → interpretation

A common failure is choosing a quantitative test for themes derived from interviews without proper coding, or collecting “preference” data but analyzing it as “cause.” Good research answers show that the analysis is justified by the type of measurement and the structure of the data.

1.4 Research ethics as part of methodology, not an afterthought

In criminology, ethics is particularly important because research may involve:

• Victims and vulnerable populations (trauma risk)
• Offenders (legal/ethical risk)
• Sensitive legal records (confidentiality)
• Police or correctional staff (power dynamics)
• Communities affected by violence (community harm risk)

CMY3702 answers are stronger when ethics are integrated into the research design, such as:

• Using anonymous reporting
• Minimizing coercion risk (no “implied authority” recruitment)
• Providing referral information if participants experience distress
• Secure storage of audio files and transcripts
• Clear consent processes (informed, voluntary, documented when feasible)

1.5 A criminology-flavoured example of coherence

Topic: Perceived police legitimacy and fear of crime among township youth
Research question: Does perceived police legitimacy predict fear of crime, controlling for age and prior victimization?
Operationalization:

• Police legitimacy → Likert scale items on trust, procedural justice, and legitimacy perceptions
• Fear of crime → Likert scale items on perceived risk and worry
• Prior victimization → self-report count or dichotomous indicator (“experienced any violent crime in past 12 months?”)
  Sampling: Survey sampling in community clusters or systematic sampling of households
  Analysis: Multiple regression or logistic regression (depending on measurement of fear)
  Interpretation: Link results to procedural justice theory; avoid claiming “police legitimacy causes fear” unless design supports causality.

This is the type of chain coherence CMY3702 expects.

1.6 South African context: design constraints and practical realities

In South Africa, criminological research often faces constraints:

• Access barriers: gatekeepers in communities, institutions, correctional facilities
• Language diversity: isiZulu, Sesotho, Afrikaans, English; need translation strategy
• Resource limitations: limited budget for large probability surveys
• Data quality issues: missing data, misunderstanding survey items
• Safety risks: fieldwork safety in high-risk areas

Methodology decisions should reflect these realities. For instance, if resources limit sampling size, students may prefer qualitative depth or smaller-sample quantitative designs—while being careful about generalization limits.

2) Research Design and Methodology Choices for Criminology at UNISA

This section builds the methodological foundation: designing studies that answer criminological questions credibly, using appropriate designs, sampling, and measurement.

2.1 Formulating research questions and objectives in criminology

A strong CMY3702-style answer often begins with a clear structure:

1. Problem statement
   • What is happening? Who is affected? Why is it important?
2. Research question(s)
   • Usually 1 main question plus supporting sub-questions
3. Research objectives
   • Typically 3–6 objectives (aligned with variables or themes)
4. Conceptual framework
   • Theoretical grounding (e.g., routine activity theory, strain theory, procedural justice theory)

Example (evaluation design):

• Problem: Community-based diversion programmes may reduce reoffending, but implementation quality differs across sites.
• Main question: How does programme implementation quality relate to reoffending outcomes among programme participants?
• Objectives:
  1. Describe variations in programme implementation quality.
  2. Measure reoffending status at a defined follow-up period.
  3. Test whether implementation quality predicts reoffending while controlling for age and prior convictions.
  4. Interpret how participants understand the programme’s impact (mixed methods).

Objectives should be action-oriented and measurable, not vague statements like “understand the programme.” If an objective says “measure,” then the study must include an operational measure.

2.2 Selecting research paradigms: positivist, interpretivist, and pragmatic approaches

Criminological research often sits between paradigms:

• Positivist / post-positivist logic (quantitative)
  • Emphasis on measurable variables, hypothesis testing, statistical inference
• Interpretivist logic (qualitative)
  • Emphasis on meaning, experiences, processes, and interpretation
• Pragmatism (mixed methods)
  • Choose methods based on the research problem and what will best answer it

In CMY3702, students should avoid claiming that “quantitative is better.” The correct approach is method fit:

• If your question is about prevalence and risk factors, quantitative is often suitable.
• If your question is about how offenders perceive rehabilitation, qualitative is often suitable.
• If you need both patterns and explanations, mixed methods can be strong.

2.3 Research designs: when to use what

2.3.1 Cross-sectional quantitative design

• Description: Data collected once at one point in time.
• Strengths: Efficient, common for baseline studies, useful for prevalence and associations.
• Limitations: Cannot establish temporal order; causality claims are weak.

Criminology example:
Measure fear of crime and perceived police legitimacy in one survey wave. You can test associations, but you cannot conclude legitimacy “causes” fear.

2.3.2 Longitudinal design

• Description: Data collected at multiple time points.
• Strengths: Better evidence of temporal relationships; can analyze change.
• Limitations: More expensive, attrition risk.

Criminology example:
Track programme participants over 12 months post-intervention to observe reoffending change.

2.3.3 Case study design (qualitative or mixed)

• Description: Deep investigation of a bounded system (a case).
• Strengths: Context-rich; supports process understanding.
• Limitations: Generalization is limited; depends on analytical rigor.

Criminology example:
Case study of how a juvenile diversion unit operates, including staff interviews and document analysis.

2.3.4 Quasi-experimental design

• Description: Intervention group and comparison group exist, but randomization may not.
• Strengths: More credible evaluation than simple before/after.
• Limitations: Confounding; selection bias.

Criminology example:
Compare reoffending rates of participants from two different sites—one with stronger programme implementation—at the same follow-up period.

2.4 Sampling strategies and sample size thinking

CMY3702 typically expects understanding of sampling logic more than memorizing a single formula.

2.4.1 Probability sampling

Used when generalization to a defined population is important.

Common types:

• Simple random sampling
• Systematic sampling
• Stratified sampling
• Cluster sampling

Criminology fit:
If you can define a sampling frame for a community and randomly select households, probability sampling strengthens inference.

2.4.2 Non-probability sampling

Used when the goal is depth, diversity of perspectives, or coverage of cases rather than statistical generalization.

Common types:

• Purposive sampling
• Snowball sampling
• Quota sampling

Criminology fit:
Interviewing offenders about rehabilitation experiences often uses purposive or snowball sampling due to access constraints.

2.4.3 Sampling for mixed methods

A typical mixed-method approach:

• Quantitative phase identifies patterns
• Qualitative follow-up explains patterns

Example:
Survey youth fear of crime. Then purposively select interview participants from survey results (e.g., high fear vs low fear) to understand why.

2.5 Measurement: operationalization and instrument quality

Measurement converts abstract concepts into data.

2.5.1 Operationalization examples for criminology constructs

1. Perceived police legitimacy
   • Items: “Police treat people fairly,” “Police enforce laws consistently,” “People should obey the police.”
2. Fear of crime
   • Items: worry about being victimized, avoiding places, perceived likelihood of harm.
3. Victimization
   • Self-reported experiences (e.g., robbery, assault) within a specified timeframe.
4. Perceived community disorder
   • Items: noise, loitering, visible public disorder signs, graffiti, vandalism.

Operationalization must specify:

• Response options (e.g., 1–5 Likert)
• Time reference (e.g., “past 12 months”)
• Direction of coding
  • Higher scores should represent the intended construct consistently.

2.5.2 Validity and reliability (quantitative emphasis)

• Validity: Are you measuring what you claim to measure?
  • Content validity (expert review), construct validity (factor structure, theory alignment), criterion validity (association with outcomes)
• Reliability: Are you getting consistent results?
  • Internal consistency (Cronbach’s alpha for Likert scales)
  • Test-retest reliability

A strong answer explains that:

• A scale can be reliable but not valid.
• Reliability often improves with clearer item wording and adequate items.

2.5.3 Translation and cultural adaptation (South Africa relevance)

In multilingual contexts:

• Translation should preserve meaning, not just words.
• Back-translation strategies can help detect inconsistencies.
• Pilot testing is essential to identify misunderstood items.

In exams, students gain marks by stating that pilot testing and translation procedures help with validity and reliability.

2.6 Data collection methods aligned to research questions

Criminology studies often use combinations of:

• Surveys (structured questionnaires)
• Semi-structured interviews
• Focus groups
• Observations (participant or non-participant)
• Document analysis (court records, policy documents, programme reports)
• Secondary data analysis (crime statistics, administrative data)

A key exam skill is matching:

• survey → variables and standardized measurement
• interviews → experiences and explanations
• documents → policies, implementation, recorded outcomes

2.7 Fieldwork management and documentation

Methodology requires transparency:

• How consent was obtained
• How interviews were recorded and transcribed
• What training fieldworkers received
• How missing data were handled
• How coding decisions were documented (for qualitative work)

In many marking schemes, poor documentation leads to reduced credibility. Students should show awareness that methodology is not only “what you do,” but “how you can justify it.”

2.8 Counter-arguments and limitations (how to score well)

Examiners often reward balanced critique. Common limitations to discuss:

• Bias
  • selection bias, social desirability bias
• Measurement error
  • inaccurate self-report, misunderstood questions
• Confounding
  • omitted variables causing spurious associations
• Ethical constraints
  • inability to access high-risk groups
• Generalization limits
  • non-probability sampling means results may not generalize

A strong CMY3702 answer doesn’t just list limitations; it connects limitation to potential effect (e.g., “social desirability may reduce reporting of deviant behaviour, underestimating prevalence”).

3) Data Analysis Foundations: From Cleaning to Interpretation (Quantitative + Qualitative)

This section covers what students typically need to demonstrate: correct analysis workflow, correct selection of tests/approaches, and correct interpretation—especially in criminological settings.

3.1 End-to-end quantitative analysis workflow

A practical, exam-friendly workflow:

1. Data management
   • codebook creation
   • variable labels
   • reverse coding for Likert scales if needed
2. Data cleaning
   • missing values handling
   • outlier checks
   • consistency checks (e.g., logical checks)
3. Descriptive analysis
   • frequencies, means, standard deviations
   • cross-tabs
4. Assumption checking
   • normality (where relevant), multicollinearity
   • independence and homoscedasticity assumptions (for linear models)
5. Inferential analysis
   • t-tests/ANOVA for group comparisons
   • correlation for association
   • regression models (linear/logistic/Poisson, as appropriate)
6. Effect interpretation
   • interpret coefficients
   • interpret odds ratios (logistic)
   • compute predicted probabilities or marginal effects where needed
7. Reporting
   • present results clearly with tables and narrative interpretation

Examiners may penalize answers that choose a test but fail to justify it or fail to interpret it in context.

3.2 Coding and data structure: the difference between nominal, ordinal, and interval

Quantitative analysis depends on measurement levels:

• Nominal variables: categories without order (e.g., gender)
• Ordinal variables: ordered categories (e.g., education levels)
• Interval/ratio variables: numeric scale where differences mean something (e.g., age)

In criminology:

• Fear of crime Likert items are often treated as ordinal, but researchers frequently build a composite scale (assuming approximate interval properties if multiple items are combined).
• Police legitimacy scales similarly can be summed/averaged into an index.

A strong answer explains decisions:

• “I will compute a composite index for perceived police legitimacy by averaging items, then treat the scale as approximately continuous for regression.”

3.3 Data cleaning in criminology: common problems and fixes

3.3.1 Missing data

Missing values can be random or systematic.

Common approaches:

• Listwise deletion (drop cases with missing values)
  • Simple but reduces sample size and can bias if missing not at random
• Mean/median imputation
  • Can distort variance and bias correlations
• Multiple imputation
  • Statistically stronger but requires more advanced processes

In exams, if the module expects conceptual understanding, students should discuss how missingness can influence validity of results.

3.3.2 Outliers

Outliers in criminology can reflect:

• true extreme cases (e.g., unusually high victimization counts)
• data entry errors
• misunderstanding of questions

A good practice:

• inspect distributions
• verify with codebook and raw entries
• do sensitivity analysis (e.g., rerun models excluding extreme errors)

3.3.3 Logical consistency checks

Examples:

• If a respondent reports “no” to victimization but then reports victimization details, that’s inconsistent.
• If age is outside plausible range, it suggests entry error.

3.4 Descriptive statistics: what to report and why

Descriptive stats are the foundation for interpretation.

3.4.1 Frequency distributions (categorical variables)

• gender distribution
• employment status categories
• prior criminal justice contact (yes/no)

3.4.2 Central tendency and dispersion (continuous or composite scales)

• mean fear of crime index
• standard deviation
• minimum and maximum scores
• distribution shape hints (skewness)

In criminology:

• If fear of crime scores are extremely skewed, parametric assumptions for regression may be impacted.
• Students should mention this risk and propose remedies (e.g., robust standard errors or alternative models).

3.5 Reliability analysis for scales (Cronbach’s alpha)

If you build a police legitimacy scale from multiple items, you should assess internal consistency.

A typical exam-ready narrative:

• Compute Cronbach’s alpha.
• If alpha is low, check item-total correlations.
• Consider removing problematic items or revising coding.
• Justify the final scale structure.

Important interpretive principle:

• High alpha does not guarantee validity, but it supports reliability.

3.6 Hypothesis testing and regression: aligning models with research questions

3.6.1 t-tests and ANOVA (group differences)

Used when comparing means between groups.

Criminology example:

• Compare mean fear of crime between two groups:
  • Group A: prior victimization = yes
  • Group B: prior victimization = no

A good answer should:

• define dependent variable (fear score)
• define grouping variable
• report group means
• report test statistic and p-value (if requested)
• interpret direction (which group is higher)

3.6.2 Correlation

Correlation is about association, not cause.

Criminology example:

• correlation between perceived police legitimacy and fear of crime

Interpretation must be cautious:

• “Negative association means higher legitimacy scores relate to lower fear scores, but it does not establish causality.”

3.6.3 Linear regression (continuous outcomes)

Used when the dependent variable is continuous.

Example model:

• Fear of crime (index) = β0 + β1*(Police legitimacy) + β2*(Age) + β3*(Prior victimization) + ε

Interpretation:

• β1 meaning: for a one-unit increase in legitimacy, fear changes by β1 units, controlling for other variables.

3.6.4 Logistic regression (binary outcomes)

When outcome is yes/no.

Example:

• Outcome: reoffending within 12 months (1=yes, 0=no)
• Predictors: programme implementation quality, age, prior convictions.

Interpretation:

• Odds ratio (OR):
  • OR < 1 indicates lower odds of reoffending associated with higher predictor values.
  • OR > 1 indicates higher odds.

3.7 Assumption checking and model validity (often tested indirectly)

Students should be aware of:

• multicollinearity (correlated predictors)
• linearity of relationships (for linear regression)
• influential points/outliers
• residual distributions

Even if a student doesn’t calculate these in an exam, they should describe why assumptions matter and what they would do if assumptions fail (e.g., transform variables, use robust methods, use non-linear models).

3.8 Example dataset scenario (consistent numbers used for illustration)

To make methods concrete, consider a small illustrative scenario (not from a real dataset). Suppose a UNISA learner conducts a survey among township youth (n=200) to examine fear of crime. A composite fear index ranges from 1 to 5 (higher = more fear). A police legitimacy index also ranges from 1 to 5 (higher = higher perceived legitimacy).

Variables:

• FearIndex (continuous)
• LegitimacyIndex (continuous)
• AgeYears (continuous, e.g., 15–24)
• PriorVictim (binary: 1 if experienced any violent crime in past 12 months)

A hypothetical regression output might show:

• β1 for LegitimacyIndex = -0.40
• β2 for AgeYears = 0.05
• β3 for PriorVictim = 0.60

Interpretation:

• A one-unit increase in perceived legitimacy (e.g., from 2.0 to 3.0) is associated with a 0.40 decrease in fear, controlling for age and victimization.
• Prior victimization is associated with 0.60 higher fear.

This example illustrates how criminological meaning must follow coefficients and remain consistent with variable coding.

3.9 Qualitative data analysis: coding, themes, and trustworthiness

Qualitative analysis is not “just reading.” It is structured meaning-making.

3.9.1 Common qualitative approaches

• Thematic analysis
  • identify codes → group codes into themes → review themes → define and name themes
• Grounded theory
  • iterative coding cycles, category development, constant comparison
• Content analysis
  • systematic categorization of textual material

In criminology, thematic analysis is common for interviews with offenders, victims, or staff.

3.9.2 A step-by-step thematic analysis workflow

1. Familiarization
   • read transcripts repeatedly
   • memo ideas
2. Initial coding
   • label meaningful segments (e.g., “fear at night,” “trust police,” “avoid streets”)
3. Searching for themes
   • group related codes (e.g., “avoidance behaviours” theme)
4. Reviewing themes
   • check coherence and whether themes reflect data
5. Defining themes
   • write theme definitions with supporting excerpts
6. Producing the report
   • integrate quotes and analysis with theory

3.9.3 Ensuring trustworthiness (qualitative validity concepts)

Key trustworthiness criteria:

• Credibility
  • triangulation, prolonged engagement, peer debriefing
• Transferability
  • thick description of context
• Dependability
  • audit trail of coding decisions
• Confirmability
  • reflexivity and documentation of researcher influence

Exams often reward students who mention:

• coding reliability steps (e.g., intercoder agreement when feasible)
• audit trails
• use of representative quotes

3.10 Integrating quantitative and qualitative findings (mixed methods logic)

Mixed methods integration can happen at different points:

• Convergent design
  • collect both types simultaneously, compare results
• Explanatory sequential design
  • quantitative first, then qualitative to explain unexpected patterns
• Exploratory sequential design
  • qualitative first, then quantitative to test or generalize themes

Criminology evaluation example:

• Quantitative results show low programme impact.
• Qualitative interviews reveal implementation barriers (e.g., low staff capacity, participant misunderstanding).
• Integration produces a richer explanation than either method alone.

3.11 Interpretation pitfalls: what examiners watch for

Common interpretation errors:

• claiming causality from cross-sectional association
• interpreting ordinal Likert outcomes as precise interval without scale justification
• “over-theming” (creating many minor themes without coherence)
• ignoring outliers or missing data patterns that distort results
• failing to connect findings to the theory or literature

Strong CMY3702 answers explicitly connect interpretation back to:

• constructs and operationalization
• the statistical meaning of coefficients
• the qualitative meaning of themes
• theoretical expectations and deviations

4) Quantitative Techniques and Reporting Practices Commonly Required in CMY3702

This section deepens quantitative technique selection and emphasizes exam-ready reporting. It also includes criminology-tailored examples.

4.1 Choosing the right analysis tool: a decision logic

A typical exam scenario: you must recommend analysis methods based on:

• measurement level of variables
• research question type (difference, association, prediction, evaluation)
• data type (cross-sectional vs longitudinal)
• sample size and distributions (at least conceptually)

4.1.1 If the outcome is continuous

• Use descriptive stats
• For predictors: correlation or linear regression
• For group comparisons: t-test (2 groups) or ANOVA (>2 groups)

4.1.2 If the outcome is binary

• Use frequencies and proportions
• For predictors: logistic regression
• For group differences in proportions: chi-square (conceptually)

4.1.3 If the outcome is count data

• Consider Poisson or negative binomial regression (conceptual expectation)
• For short answers: discuss appropriateness and distribution concerns

4.2 Correlation versus regression in criminology narratives

Students often confuse the difference. A crisp exam-quality distinction:

• Correlation: describes association between two variables.
• Regression: models the relationship between an outcome and one or more predictors while controlling for other variables.

In criminology:

• Correlation between police legitimacy and fear might reflect underlying differences in age distribution.
• Regression helps control for age and prior victimization.

4.3 Multiple regression: controlling confounding

Criminology questions often require multiple predictors:

• fear of crime depends on legitimacy, victimization, neighbourhood disorder, personal circumstances.

A multiple regression approach:

• reduces omitted variable bias (not eliminates it)
• provides conditional effects

Exam narrative example:

• “After controlling for prior victimization and age, the association between legitimacy and fear remains negative.”

4.4 Logistic regression: odds ratios and interpretive accuracy

Students can lose marks by misinterpreting odds ratios.

Correct conceptual interpretation:

• Odds ratio (OR) > 1 indicates increased odds of outcome for higher predictor values.
• OR < 1 indicates decreased odds.
• OR = 1 implies no association.

Example narrative:

• OR = 0.60 for a one-unit increase in programme participation indicates lower odds of reoffending.

If predictor is continuous, interpret per unit change. If predictor is coded as categories (e.g., high vs low implementation quality), interpret relative comparisons.

4.5 Chi-square tests for categorical variables (conceptual reporting)

A typical use:

• association between category variables:
  • prior victimization (yes/no) and fear category (low/medium/high; if categorized)
• report:
  • chi-square statistic, degrees of freedom, p-value

Interpretation:

• if p < 0.05, conclude evidence of association (but not causality).

4.6 One-way and two-way ANOVA: group comparisons beyond t-tests

When groups >2:

• One-way ANOVA compares one categorical independent variable with a continuous dependent variable.

Two-way ANOVA extends:

• compares effect of two categorical independent variables and interaction.

Criminology example:

• FearIndex by:
  • prior victimization (yes/no)
  • policing satisfaction (low/high)

Interaction means:

• effect of victimization differs by policing satisfaction level.

In exams, if the question is limited to one-way comparisons, students should not assume two-way unless asked.

4.7 Practical reporting: how to write results sections

In UNISA-style assignments/exams, results should be structured:

1. Sample description
   • n, key demographic distributions
2. Scale quality
   • reliability (alpha) for indices
3. Descriptive outcomes
   • mean fear, distribution, key categories
4. Inferential results
   • test statistics, p-values, effect sizes if available
5. Interpretation
   • connect to research question and theory

A good results paragraph is not just numbers; it interprets the relationship.

4.8 Effect sizes: why they matter

P-values can mislead if sample sizes are large or small.

Students should aim to report:

• standardized coefficients (where relevant)
• odds ratios (logistic)
• R-squared (linear regression)
• or at least Cohen’s d for t-tests/ANOVA if asked

Even if an exam question does not request effect size calculation, discussing effect size as important earns marks.

4.9 Handling categorical predictors and reference categories (logistic regression)

In logistic regression:

• one category is reference (baseline)
• coefficients show changes relative to reference

Example:

• Programme implementation quality coded:
  • Low (reference)
  • Moderate
  • High

Interpretation example:

• OR for High vs Low > 1 means higher odds; OR < 1 means lower odds.

Students should state reference group explicitly in interpretation.

4.10 Assumptions and robustness: exam-level expectations

If asked about assumptions, students should not panic. They can describe:

• Linear regression
  • linearity
  • normality of residuals (approximate)
  • homoscedasticity
  • independence
• Logistic regression
  • linearity of logit for continuous predictors (conceptually)
  • independence of observations
  • sufficient sample size

If assumptions violated:

• propose robust alternatives:
  • transform variables
  • use robust standard errors
  • choose alternative models (e.g., ordinal logistic regression if appropriate)

4.11 A criminology-tailored quantitative mini-case (with coherent interpretation)

Scenario: A study examines whether youth perceptions of procedural justice reduce fear of crime.
Variables:

• FearIndex (1–5)
• ProceduralJusticeIndex (1–5)
• CommunityDisorderIndex (1–5)
• PriorVictim (0/1)

Hypothetical findings:

• ProceduralJusticeIndex coefficient: -0.30 (p < 0.01)
• CommunityDisorderIndex coefficient: +0.45 (p < 0.001)
• PriorVictim coefficient: +0.55 (p < 0.01)

Interpretation:

• Higher procedural justice is associated with lower fear, even when disorder and victimization are controlled.
• Disorder increases fear strongly.
• Prior victimization increases fear.

This is the kind of interpretation that stays criminologically meaningful: justice reduces fear; disorder and victimization increase fear.

4.12 Counter-arguments and alternative explanations (high-value exam content)

Even with significant results, alternative explanations exist:

• Reverse causality: those with higher fear may perceive police as less legitimate.
• Omitted variable bias: neighbourhood media exposure could influence both legitimacy perceptions and fear.
• Measurement bias: fear reporting might be affected by social desirability or misunderstanding.

A strong answer acknowledges these limitations and proposes improved designs:

• longitudinal follow-up
• cross-lagged analysis (advanced)
• instrument triangulation and pilot testing
• quasi-experimental evaluation for programme impacts

5) Qualitative Methods, Mixed Methods, and Full Research Output Skills (Writing, Defending, and Analysing)

This section focuses on qualitative methodology and the skills needed to produce a coherent CMY3702 research output: combining methods, defending analytical decisions, and producing a credible final report.

5.1 Why qualitative research is crucial in criminology

Quantitative studies can show patterns; qualitative studies explain meaning and mechanisms. In criminology, this matters because behaviour is often shaped by:

• perceived legitimacy and trust
• cultural norms
• fear and avoidance routines
• experiences with policing and courts
• institutional constraints and stigma
• programmatic experiences (what participants believe is helping them)

5.2 Designing qualitative studies: interviews, sampling, and credibility

5.2.1 Choosing qualitative sample logic

Qualitative sampling aims for:

• diversity of experiences
• theoretical relevance
• data saturation (when no new themes emerge)

Purposive sampling strategies:

• maximum variation: recruit participants from different contexts
• homogeneous sampling: focus on one group to deep explore
• typical case sampling: focus on “average” experiences

Criminology example:

• Interviews with youth who have low fear and youth with high fear of crime to compare meanings (maximum variation strategy).

5.2.2 Building an interview guide

An interview guide should:

• align with research questions
• use clear, non-leading questions
• include probes to explore depth

Example question set for procedural justice and fear:

• Main: “Tell me about a time you interacted with the police. How did that experience make you feel?”
• Probe: “What actions by the police mattered most?”
• Main: “How does that influence how you move around your community?”
• Probe: “What do you do differently when you feel unsafe?”

Interview guides in CMY3702 answers often gain marks when:

• questions are tied to constructs
• probes are included
• language is appropriate for the community
• the interviewer is trained to manage sensitive topics ethically

5.3 Qualitative analysis in more depth: coding strategies and example theme development

5.3.1 Developing codes and codebooks

Coding can be:

• deductive (from theory)
• inductive (emerging from data)
• hybrid

A codebook should include:

• code name
• definition
• inclusion/exclusion criteria
• example quote(s)

Example (police legitimacy):

• Code: “Perceived Fair Treatment”
  • Definition: statements describing fairness in how police treat community members
  • Inclusion: mentions respect, equal treatment, not abusing authority
  • Exclusion: general fear without fairness reference

5.3.2 Moving from codes to themes

A theme is a higher-level pattern that explains or organizes multiple codes.

Example theme development:

• Codes: “Respectful dialogue,” “Explained reason for stopping,” “No discrimination”
• Theme: “Procedural justice cues build perceived legitimacy”

Another theme:

• Codes: “Avoiding certain streets,” “Staying indoors at night,” “Planning routes”
• Theme: “Fear shapes everyday mobility and avoidance”

5.3.3 Saturation and evidence sufficiency

Students can mention:

• saturation is reached when additional interviews do not contribute new codes/themes.
• criteria can be documented through memoing.

5.4 Mixed methods integration: practical and defensible ways

Mixed methods can integrate:

• through design (convergent vs sequential)
• through analysis (joint displays)
• through interpretation (connecting mechanisms to patterns)

5.4.1 Joint displays as integration tools (exam-friendly concept)

A joint display table can align:

• quantitative results (e.g., fear mean differences)
• qualitative themes (e.g., why fear increases)

Even if you do not create a table in an exam, you can describe the logic:

• “Where the survey shows higher fear among prior victimization youth, interviews suggest that victimization creates learning about risks and mistrust in protective responses.”

5.5 Building and defending a conceptual framework

CMY3702 often benefits when students link to criminological theory.

Examples of theory-function mapping:

• Routine activity theory → explains victimization risk (presence of motivated offenders, absence of guardianship)
• Procedural justice theory → explains legitimacy and compliance, potentially affecting fear and cooperation
• Strain theory → explains how blocked goals can increase deviant coping
• Social learning theory → explains peer influence and criminal behaviour through reinforcement and modeling

Defending the framework means:

• your operationalization and interview questions are consistent with theoretical constructs
• your analysis interprets findings through those constructs

5.6 Writing a research report: structure that earns marks

A standard criminological research report structure:

1. Title and abstract
2. Introduction
   • context, problem statement, rationale, research question/objectives
3. Literature review
   • theory and prior findings
4. Methodology
   • design, sampling, data collection, instruments, analysis plan, ethics
5. Results/findings
   • quantitative tables and qualitative themes with excerpts
6. Discussion
   • interpret results, connect to theory and literature, acknowledge limitations
7. Conclusion and recommendations
   • answer question, implications for policy/practice
8. References and appendices
   • instruments, consent forms, coding framework (if required)

In exams, even short answers can reflect this structure by:

• presenting “Method” details before “Results”
• linking discussion to research question and objectives
• concluding with direct answers and justified recommendations

5.7 Ethics and positionality in qualitative and mixed methods

In qualitative work, researchers may have:

• insider/outsider positionality
• power dynamics with participants
• emotional load (hearing trauma narratives)

Ethical defensibility includes:

• reflexive memoing
• interviewer training
• safe interview environments
• managing distress and providing support pathways
• maintaining confidentiality in reporting (avoid identifiable details)

In exam responses, “positionality” should be mentioned briefly but specifically:

• how the researcher’s identity could influence responses
• what steps reduce harm or bias

5.8 Quality assurance in research: reliability, validity, and audit trails

Quality in mixed methods requires:

• quantitative reliability (scale alpha, consistent coding)
• qualitative trustworthiness (credibility, confirmability)
• integration clarity (how results are compared/connected)

A defensible audit trail includes:

• how data were coded
• changes to the coding framework
• how decisions about missing data were made
• how analysis steps were documented

5.9 Common exam questions and how to answer them well (templates)

5.9.1 “Describe and justify a research design”

Strong answers include:

• identify design type (cross-sectional, longitudinal, case study, quasi-experimental)
• state why it fits the question
• state strengths and key limitations
• propose ethical procedures relevant to the design

5.9.2 “Explain sampling and bias”

A strong answer includes:

• sampling method and reason
• how sample relates to target population
• bias types (selection, social desirability)
• mitigation strategies (training, careful recruitment procedures)

5.9.3 “Discuss validity and reliability of a measurement instrument”

Strong answers:

• define validity and reliability in your own words
• state how you would test them (pilot testing, alpha)
• mention translation and cultural adaptation in South Africa context
• address measurement limitations

5.9.4 “Explain data analysis steps”

Strong answers:

• show sequence (clean → descriptive → assumptions → inferential → reporting)
• align test choice with variable type
• interpret coefficients and link back to research question

5.9.5 “Conduct thematic analysis: show coding to themes”

Strong answers:

• provide example codes and one theme
• include a conceptual definition of the theme
• show how codes logically cluster
• mention trustworthiness strategies

5.10 A full integrated example: UNISA-style criminology mixed methods proposal (coherent end-to-end)

Topic: Programme implementation quality and reoffending outcomes among diversion programme participants
Main research question: Does stronger programme implementation predict lower reoffending at 12 months, and how do participants explain what helps reduce reoffending?
Objectives:

1. Measure programme implementation quality across sites.
2. Determine reoffending status at 12 months post-programme.
3. Test whether implementation quality predicts reoffending controlling for age and prior convictions.
4. Explore participants’ experiences to explain mechanisms linking implementation to outcomes.

5.10.1 Design choice

• Explanatory sequential mixed methods
  • Phase 1: quantitative association/prediction
  • Phase 2: qualitative explanation of mechanisms

5.10.2 Sampling

• Quantitative:
  • participants recruited from programme sites using site-based recruitment lists
  • aim for adequate sample size for logistic regression (conceptually)
• Qualitative:
  • purposive sampling from quantitative results
  • select participants from “low reoffending” and “high reoffending” groups to compare experiences

5.10.3 Measurement

• Implementation quality index:
  • staff training adequacy, session adherence, participant engagement measures (constructed index)
• Reoffending outcome:
  • binary: reoffended within 12 months (yes/no)
• Controls:
  • AgeYears
  • PriorConvictions (e.g., 0/1 or count)

5.10.4 Quantitative analysis

• descriptive statistics: implementation quality distribution, reoffending proportion
• reliability analysis for implementation index (Cronbach’s alpha if Likert-based)
• logistic regression:
  • outcome: reoffended (0/1)
  • predictors: implementation quality + age + prior convictions

Interpretation:

• If higher implementation quality reduces odds of reoffending, odds ratios < 1 would be expected.

5.10.5 Qualitative analysis

• interview participants about:
  • perceived helpfulness of programme sessions
  • how staff interactions influenced behaviour change
  • barriers to participation and motivation
• thematic analysis:
  • codes → themes (e.g., “consistent support,” “accountability experiences,” “missed sessions undermining change”)

5.10.6 Integration

• compare quantitative pattern (which sites or levels reduce reoffending) with qualitative mechanisms:
  • Example linkage:
    • Quantitative suggests higher implementation quality predicts lower reoffending.
    • Qualitative explains: participants feel consistent guidance, accountability, and follow-up.

5.10.7 Ethics

• informed consent
• confidentiality of justice outcomes
• secure data storage
• referral options if discussing reoffending triggers distress

This integrated example demonstrates methodological coherence across research question, sampling, measurement, analysis, interpretation, and ethics—exactly the kind of alignment that strengthens CMY3702 submissions.

5.11 Limitations and recommendations: how to conclude strongly

A strong discussion section addresses:

• whether results match theory or contradict it
• plausible reasons for deviations
• limitations (design constraints, measurement limits, missingness)
• implications:
  • for policy (programme funding or training standards)
  • for practice (improving fidelity of implementation)
  • for future research (longitudinal evaluation, stronger causal designs)

Recommendations should be grounded:

• if analysis shows implementation quality matters, recommend focusing on implementation fidelity
• if qualitative themes show barriers, recommend addressing those barriers operationally (e.g., scheduling, participant support)

Final synthesis: the “mark-winning” CMY3702 approach

To perform exceptionally in UNISA CMY3702 Criminological Research: Methodology and Data Analysis, students should consistently demonstrate:

• Alignment: research question → design → sampling → measurement → analysis → interpretation
• Ethics integration: ethical planning is part of methodology
• Measurement competence: operationalize constructs clearly and justify indices/scales
• Analytical correctness: choose methods based on variable types and research aims
• Interpretive responsibility: report results accurately and avoid causal overreach
• Credible qualitative rigor: systematic coding, defensible themes, and trustworthiness
• Strong reporting: clear structure, coherent narrative, and evidence-based conclusions

These notes emphasize not only what to do, but how to defend why it was done—the skill at the core of criminological research methodology and data analysis.