Tampere University Public and Global Health Master's Thesis Guide: 30 ECTS, PGH.502 and Trepo

Quick answer: what is the current Public and Global Health thesis route?

Tampere University’s Public and Global Health (PGH) is a 120 ECTS Master of Health Sciences programme. In the current 2026-2027 curriculum, the degree explicitly includes a 30 ECTS master’s thesis within Advanced Studies. PGH.502 Master’s Thesis Seminar is a separate 5 ECTS pass/fail course, so the thesis-study pathway totals 35 ECTS while the thesis itself remains 30 ECTS and is graded 0-5. The current applicant page says No intake in 2026; that statement should not be stretched into assumptions about later admission years.

1. Read the current credit structure, not an older PGH curriculum

The current degree structure is Joint Studies 7-12 ECTS, Advanced Studies 95-110 ECTS and Free Choice Studies 3-18 ECTS. PGHM.AS-S02 is the current advanced-studies module and says compulsory core courses plus Master Thesis Studies total 85 ECTS. Older Tampere material may show a 25 ECTS PGH thesis. Do not use that historical value for the 2026-2027 guide: the current degree says 30 ECTS. Likewise, do not call the whole 35 ECTS pathway a 35 ECTS thesis; five of those credits belong to PGH.502.

2. PGH.502 is a real thesis seminar, not an administrative placeholder

PGH.502 Master’s Thesis Seminar is 5 ECTS, C1 level and pass/fail. The seminar introduces the thesis process and structure, requires a research plan, progress presentations and critical reading/commenting of peers’ work, and covers ethics of scientific writing including plagiarism. Current completion information states compulsory attendance and all parts compulsory. Use the seminar to test whether your question is answerable, whether your methods actually identify the intended population/exposure/outcome, and whether your evidence claims survive critique before the final analysis is locked.

3. The thesis is 30 ECTS and graded 0-5

The current PGH degree independently states a 30 ECTS Master’s Thesis. Tampere’s general non-technical thesis framework grades master’s theses 0-5. Keep this separate from PGH.502, which is pass/fail. The thesis may be empirical or based on existing research literature, but it must demonstrate independent scholarly work and a defensible research process. A good PGH thesis therefore needs a bounded public/global health question, a transparent evidence chain, correct ethical/data permissions and conclusions that match the design rather than the social importance of the topic.

4. Public and Global Health is deliberately interdisciplinary

PGH combines health sciences and social sciences. The current programme spans epidemiology, health promotion, health policy, health protection and security, health systems, health economics, social determinants, social protection, environment, sustainable food systems and global health governance. That breadth is useful for topic selection but dangerous for scope. A 30 ECTS thesis should not attempt to solve “health inequality” or “global health governance” in general. Choose one population, policy, exposure, service, system level, mechanism or evidence gap that can be analysed rigorously.

5. Start with the unit of analysis and population

Before choosing software or a theoretical framework, state the unit of analysis: person, household, organisation, municipality, region, country, policy document, health-system unit or another clearly defined object. Then define the target population and the population actually observed. In public health, denominators are part of the argument. A percentage without a clear denominator, a national indicator without a clear year/population definition, or an interview sample without a clear recruitment frame makes later interpretation unstable.

6. Distinguish prevalence, incidence, association and causality

Current PGH.201 Basic Epidemiology explicitly covers measures of occurrence and association, study designs and causality. Use those distinctions. Prevalence describes existing cases or states at a defined time/period; incidence concerns new occurrence over time; association describes statistical dependence; causal effect requires stronger design and assumptions. Do not write that an exposure “caused” an outcome merely because a cross-sectional regression coefficient is statistically significant. State what the design can and cannot identify.

7. Operationalise exposure and outcome before analysing them

PGH.201 also emphasises defining exposures and outcomes and measurement validity. Write an operational definition table before analysis: variable, source, time window, units/categories, coding, missing values and role in the model. If you derive an exposure from several survey items or classify a health outcome using a threshold, justify the rule. Changing definitions after seeing results may be legitimate in sensitivity analysis, but it should be disclosed rather than presented as if it had been pre-specified.

8. Bias and confounding need a design-specific audit

Systematic error, random error and confounding are current PGH.201 core content. For your actual design, ask how participants entered the dataset, how exposure/outcome were measured, what was not observed and which common causes could distort an association. Discuss selection bias, information bias and confounding in relation to the specific study, not as a generic limitations list. If you adjust for covariates, explain why those variables belong in the model instead of describing adjustment as an automatic route to causal truth.

9. Cross-sectional, cohort, case-control and ecological evidence have different limits

A cross-sectional study usually cannot establish whether exposure preceded outcome. A cohort design gives temporal ordering but can still suffer confounding, selection and measurement problems. Case-control studies require careful source-population and exposure reasoning. Ecological comparisons operate at group level and create an ecological-fallacy risk if interpreted as individual effects. The thesis should explicitly connect design choice to the research question and phrase conclusions at the level that design supports.

10. Surveys need a denominator, recruitment story and nonresponse analysis

For a survey thesis, define the target population, sampling frame or recruitment channel, invitation process, inclusion/exclusion criteria and final analytical denominator. Explain whether participation was probabilistic, convenience-based, workplace/school-based, online open recruitment or another mechanism. If response rate or nonresponse patterns can be estimated, discuss how they may affect representativeness. Weighting does not automatically repair every selection problem; state what population feature the weights are intended to recover and how they enter the analysis.

11. Secondary and registry data require provenance

Existing data can be powerful and efficient, and current PGH guidance recognises thesis work using existing or self-collected data. But secondary data are not method-free. Record dataset/registry name, version or extraction date, inclusion period, target population, original collection purpose, variable definitions, coding changes, linkage/derivation and your analytical exclusions. If health data are pseudonymised, they can still be personal data when individuals remain linkable. Permissions and data-protection duties do not disappear because somebody else collected the data first.

12. Missing data must be visible

Report missingness by important variable and, where useful, by key groups. State whether complete-case analysis, explicit missing categories, imputation or another approach was used and what assumptions it requires. A model run on fewer observations than the descriptive table should not silently change the study population. If missingness may relate to exposure, outcome or participation, discuss the direction in which it could bias interpretation. Sensitivity analyses are more useful when they test a plausible missing-data concern rather than simply generating extra tables.

13. Statistical models should answer the question, not showcase software

Current PGH.304 covers ANOVA, linear regression and logistic regression, along with interpretation of statistical results. Select a model based on outcome type, design and estimand. Define reference groups, coding, transformations and interaction terms. Check relevant assumptions and diagnostics. Do not confuse odds ratios with risk ratios. Report effect estimates with uncertainty, such as confidence intervals when appropriate, and distinguish statistical detectability from public-health importance.

14. Covariates, subgroup analysis and multiplicity need a plan

A long covariate list is not automatically a better model. Justify adjustment using subject-matter reasoning and the causal or descriptive purpose of the analysis. Avoid adjusting for variables that may lie on the causal pathway unless that matches the estimand. Predefine key subgroup analyses when possible, report denominators and uncertainty, and avoid treating an effect significant in one subgroup but not another as proof that the subgroups differ. Label extensive outcome/model exploration as exploratory.

15. Qualitative PGH work needs transparent sampling and analysis

Current PGH.203 Qualitative health research covers data gathering, ethics, thematic analysis and thorough reporting. Explain why particular participants, organisations, documents or settings were selected. Document recruitment, consent, interview/focus-group setting, recording/transcription and the analytic process. Preserve a traceable path from raw material to codes, themes and final claims. Qualitative rigour comes from coherent design and transparent interpretation, not from using a fashionable method label.

16. Reflexivity and sample sufficiency matter

In public/global health, access and positionality can shape whose voices are heard and how accounts are interpreted. Describe relevant researcher roles, relationships, language, professional background or institutional access. If you use “saturation,” explain what kind of saturation or sample sufficiency you mean and how you judged it. A small purposeful sample can be appropriate for a focused qualitative question, but the thesis must not quietly generalise it as a prevalence estimate for a population.

17. Translation and multilingual material need an audit trail

Global-health research often crosses languages. If interviews, policy documents or quotations are translated, state who translated them, at what stage and how meaning-sensitive decisions were checked. Preserve original-language text where permitted for analytic verification. Translation can change conceptual nuance, especially for health-system, stigma, gender, policy or culturally embedded terms. Do not present translated words as perfectly equivalent without acknowledging interpretive choices when those choices affect the findings.

18. Mixed methods requires integration, not just two datasets

PGH explicitly supports quantitative and qualitative research. A mixed-methods thesis should explain why both are needed and where integration occurs. The qualitative strand might explain mechanisms behind a quantitative pattern, or quantitative data might test the distribution of a theme found qualitatively. Two parallel analyses with separate conclusions are not automatically mixed-methods research. Plan the point of integration, identify contradictions and explain whether one strand changes interpretation of the other.

19. Global-health comparisons need context and comparability

Current PGH global-health modules emphasise institutions, law, policy, environment, social protection and inequalities from global to local levels. In cross-country work, define the policy period, data source, health-system context and indicator comparability. Countries are not exchangeable experimental units. Differences may reflect measurement systems, population structure, welfare institutions, conflict, reporting capacity or policy timing. Avoid deficit framing that treats one setting simply as a failed version of another.

20. Policy text, implementation and health outcome are separate evidence levels

A law, strategy or WHO recommendation shows that a policy position exists; it does not prove implementation. Implementation activity does not automatically prove population health impact. For policy-document analysis, record issuing body, authority, date/version, intended population and implementation status. If you evaluate Health in All Policies or another cross-sector strategy, identify the non-health policy domain, proposed health pathway, governance level and evidence that the policy was actually enacted or affected relevant processes.

21. Use organisational and grey literature critically

WHO reports, government documents, NGO publications and programme evaluations can be essential primary sources for global/public health, especially for policy and implementation questions. Treat them according to their role. Ask who produced the document, for what purpose, which data underpin it, how methods are described and whether it makes advocacy, monitoring or causal claims. Do not flatten all sources into one evidence tier or assume that institutional authority converts a descriptive report into causal evidence.

22. Health systems and health economics require correct units and methods

For health-system research, state whether the unit is a facility, service, municipality, region or national system and which system function is being studied: governance, financing, workforce, access, quality or another defined component. For health economics, costs alone do not constitute cost-effectiveness. Economic evaluation needs a defined comparator, perspective, time horizon, outcomes and analytic framework appropriate to the question. If your thesis only describes expenditure or affordability, call it that.

23. Social determinants and equity need structural interpretation

Current PGH teaching explicitly addresses poverty, social protection and social determinants of health. Define the stratifier-income, education, gender, migration status, occupation, region or another variable-and the mechanism you are evaluating. Distinguish individual-level associations from structural explanations. Where possible, report both absolute and relative differences because they can tell different equity stories. Avoid implying that disadvantaged populations cause their own risk when the evidence concerns institutional or material conditions.

24. Environment and climate-health studies need aligned space and time

Environmental exposures and health outcomes can operate at different spatial and temporal scales. Define exposure resolution, outcome period, lag assumptions and geographic linkage. A country-average climate indicator paired with individual health data requires careful inference. Spatial clustering and temporal trends can create apparent associations. Keep ecological evidence at the ecological level unless a design genuinely links exposure to individuals. Clearly separate observed associations from projections or scenario-based estimates.

25. Programme evaluation should distinguish process, output, outcome and impact

A programme may reach participants, deliver activities and be well accepted without changing health outcomes. Define the evaluation question: implementation fidelity, reach, feasibility, acceptability, service outcome, behaviour, health endpoint or longer-term impact. Pre/post improvement alone does not prove the programme caused the change if secular trends, selection or regression to the mean are plausible. State the counterfactual limitation and avoid escalating implementation success into effectiveness without evidence.

26. Sensitive and vulnerable populations require governance before data collection

Public/global health research may involve sensitive diagnoses, migration status, violence, poverty, children, marginalised groups or other vulnerabilities. Work with supervisors to determine the applicable ethics review, organisational permissions, consent/information process and data-protection roles before collection or use. Design recruitment and reporting to reduce re-identification and harm. The public nature of the final thesis makes disclosure planning especially important for small communities, rare conditions or politically sensitive contexts.

27. Personal health data need minimisation and lifecycle planning

Before receiving or collecting personal data, define what variables are necessary, where the data will be stored, who has access, how identifiers/keys are separated, what is retained and what enters the public thesis. Pseudonymisation reduces exposure but does not make linkable data anonymous. For secondary datasets, document the permission basis and processing role. Tables and quotations should also be disclosure-checked; combining several harmless-looking variables can re-identify participants in small samples.

28. AI use must respect evidence and confidentiality

Follow Tampere’s current AI guidance and your supervisor’s thesis-specific rules. AI can help with brainstorming, coding explanations, language or debugging, but the student remains responsible for citations, statistics, translations and interpretations. Verify generated references against primary sources and rerun generated code on controlled data. Do not upload protected participant, health-system, partner or unpublished research data to external AI services without an approved basis.

29. Build a reproducibility record while the study is running

Maintain a research log linking dataset/version, inclusion flow, variable dictionary, data-cleaning decisions, analysis script, model version and final table/figure. For qualitative work, preserve the codebook, memo trail and theme-development decisions consistent with permissions. For policy/document work, archive document versions and retrieval dates. A final thesis is much easier to defend when every reported number or claim can be traced back to a defined source and analytical decision.

30. Supervision and PGH.502 should be used as methodological gates

Bring the research question, design, variable definitions or sampling plan to supervision early. Use PGH.502 presentations to expose mismatches before they become expensive: a causal question with cross-sectional data, a national claim from a local convenience sample, a mixed-methods design without integration, or a policy-impact question with only policy texts. Record major scope or method changes and why they were made. A narrower question answered rigorously is usually stronger than a broad one supported by ambiguous evidence.

31. Maturity, Turnitin, Trepo and public-thesis rules are separate gates

The non-technical master’s thesis process includes the maturity requirement, originality review and repository submission. Follow current Tampere instructions for the applicable maturity route. Complete the Turnitin originality workflow as required, prepare the final file in the required archive-compatible format such as PDF/A, and submit through Trepo. The assessed thesis is public, so keep confidential participant, employer or partner material outside the public manuscript when required rather than assuming a thesis can simply be made secret.

32. Plan the examiner and graduation timeline backwards

Tampere’s current general thesis guidance gives examiners 21 days, extended to 28 days if a separate maturity test must also be completed as part of the thesis process. This is not the entire graduation timeline. Allow time for supervisor review, revisions, seminar completion, maturity, Turnitin, Trepo processing and faculty decisions. Work backwards from the desired graduation date and check current faculty deadlines rather than submitting the thesis on the last possible day.

33. Run an epidemiology and bias audit before freezing the Results

Before finalising the quantitative Results, rerun the logic of the epidemiology rather than only rerunning the software. Check the target population, denominator, exposure and outcome definitions, temporality, selection process and missing-data flow. Then list the most plausible sources of bias and confounding and ask whether the model actually addresses them. Statistical adjustment cannot repair every form of selection or measurement bias, and a precise confidence interval can surround a biased epidemiology estimate. If a result changes under a justified sensitivity analysis, explain what assumption the sensitivity analysis tested. If causal language remains in the Discussion, verify that the design, temporality and confounding assumptions support it. This epidemiology audit should leave the reader able to distinguish descriptive epidemiology, analytic epidemiology, epidemiologic association and any genuinely defensible causal interpretation.

34. Run a qualitative and global-health interpretation audit

For qualitative or mixed-methods work, trace several final claims backwards through themes/codes to the underlying material and check that negative or deviant cases were not hidden. Re-read the reflexivity, translation and sample-sufficiency statements and make sure they describe what actually happened. For global health and global-health policy research, audit the country/time context, policy version, implementation status, actor authority and source purpose. Separate global-health framing, human-rights arguments, policy adoption, implementation evidence and observed health outcomes. In comparative global-health work, ask whether indicator definitions and data coverage are genuinely comparable. If the thesis combines qualitative evidence with epidemiology or statistics, show exactly where the strands change or strengthen the interpretation rather than simply placing two sets of findings next to each other.

33. Final Public and Global Health checklist

Before submission verify: 30 ECTS thesis, PGH.502 5 ECTS pass/fail seminar, current Sisu structure, clearly defined population/unit/denominator, design matched to the research question, exposure/outcome definitions fixed, bias/confounding addressed, missing data visible, model assumptions checked, qualitative sampling/coding/reflexivity traceable, mixed-methods integration explicit, policy/implementation/outcome levels separated, equity/global context interpreted carefully, ethics and personal-data permissions resolved, AI use compliant, public/confidential material separated, maturity/Turnitin complete, PDF/A valid, Trepo submitted and examiner time reserved. A strong PGH thesis makes an important health question more precise rather than making its evidence sound more certain than it is.