Level of analysis determines whether a public health nursing dissertation examines people, households, teams, services or areas. Align the research question, observations and intended conclusions, paying particular attention to repeated contacts, clustering and the difference between individual and area-level relationships.
This guide examines the level at which the dissertation can make a defensible claim. Its practical output is a unit-of-observation and unit-of-inference map distinguishing people, episodes, teams and areas. Begin with the how to keep a public health dissertation focused on nursing if you first need to establish the broader project approach. Use the detailed decisions below once that foundation is clear.
Understanding Levels of Analysis
In public health nursing, the level of analysis denotes the specific focus of your research inquiry. A clear grasp of these levels ensures that you address the right questions and utilise data effectively. The primary levels include:
- Person: This level involves individual assessments and interventions.
- Household: Focusing on family dynamics and health behaviour.
- Nurse/Team: Analysing the roles and functions of healthcare teams.
- Service: Evaluating the effectiveness of health services.
- Geographical Area: Assessing health trends and outcomes in specific regions.
Utilising the correct level of analysis helps avoid ecological and atomistic inference errors, where conclusions drawn from data at one level do not accurately represent that at another.
Common Errors in Analysis
When selecting your level of analysis, be wary of common errors. Ecological inference errors occur when data from groups are used to make claims about individuals. Conversely, atomistic inference errors do the opposite, making inferences about groups based solely on individual data. It’s critical to align your research question and available data to prevent these issues from cropping up in your dissertation.
Illustrative Example 1: Person-Level Analysis
Consider a fictional study aimed at exploring the mental health outcomes of adolescents in a particular urban area. A nursing student decides to conduct surveys with individuals aged 13-18. This person-level analysis allows the student to gain insights into specific variables affecting mental health, such as socioeconomic background and peer relationships.
The fictional student proposes interviews about individual experiences and explains how the recruitment context may limit interpretation. Any association found would need an appropriate analysis and cautious reporting. The example does not provide real findings or establish a statistically significant relationship.
Illustrative Example 2: Household-Level Analysis
Another student might focus on a household-level analysis to examine health behaviours in low-income families. By interviewing family members collectively, the student discovers patterns in health behaviours and access to resources.
For instance, families that engage in shared meal planning exhibit healthier eating patterns compared to those who do not. The student presents these findings to propose community workshops aimed at improving family nutrition, illustrating the importance of family dynamics in health promotion.
Choosing the Right Level of Analysis
Selecting the correct level of analysis is not merely academic; it has practical implications for public health interventions. To choose wisely, consider the following:
- Research Question: Define what you aim to discover.
- Data Availability: Ensure data at the chosen level is accessible.
- Feasibility: Evaluate the resources and time required for your analysis.
Comparison Table of Levels of Analysis
| Level of Analysis | Focus Area | Data Collection Methods | Example Questions |
|---|---|---|---|
| Person | Individual behaviour | Interviews, Surveys | How does stress impact individual health? |
| Household | Family dynamics | Interviews, Observations | What behaviours contribute to family health? |
| Nurse/Team | Healthcare delivery | Observations, Case Studies | How do team dynamics affect patient outcomes? |
| Service | Service effectiveness | Surveys, Performance Metrics | How effective is a local health initiative? |
| Geographical Area | Population health | Statistical Records, Surveys | What health trends exist in this community? |
Aligning Questions and Data
When crafting your dissertation, ensure your research questions are well-aligned with available data. This synergy is crucial in obtaining meaningful results. An analysis that doesn’t take into account the level of analysis may lead you to conclusions that are either unsubstantiated or irrelevant.
Step-by-Step Checklist for Selecting Your Level of Analysis
- Identify your research interests: What topics excite you within public health nursing?
- Define your research question: What specific inquiry do you wish to explore?
- Evaluate available data: What data can you access at the chosen level (individual, household, etc.)?
- Reflect on the feasibility: Do you have sufficient resources to carry out this analysis?
- Identify potential biases: Are there any biases in your data collection method?
- Consider ethical guidelines: Does your study align with ethical research practices?
- Refine your methodology: Based on the previous answers, refine your selection to the most suitable level of analysis.
Ethical Considerations in Research
When conducting research in public health nursing, adhering to ethical guidelines is paramount. The Health Research Authority (2026) provides a comprehensive framework for maintaining integrity during research, outlining responsibilities such as obtaining informed consent, ensuring confidentiality, and minimising harm to participants.
Level of analysis: start with the claim you want to make
Write one sentence describing the conclusion your dissertation is intended to support. Identify its subject: an individual patient, a household, a nursing team, a service or a geographical area. That subject is your first clue about the appropriate level of analysis. Next, identify the observations you actually have. A dataset containing one row for each local authority cannot directly answer a question about differences between individual patients, even when the variables have familiar clinical names.
This distinction concerns the structure of the evidence. It does not mean that one level is more valuable than another. Area data can answer useful questions about variation between places. Patient interviews can illuminate individual experience. Team records can describe how services organise work. The problem arises when an analysis answers one question and the discussion quietly claims to answer a different one.
Separate observation, intervention and inference
The unit of observation is what each recorded observation represents. The intervention may operate somewhere else: a training programme could be delivered to teams while outcomes are recorded for patients. The unit of inference is the entity about which you intend to draw conclusions. Record all three explicitly. If they differ, explain how your design and interpretation account for that difference rather than forcing them into a single label.
For example, a fictional evaluation might train four community nursing teams and collect experience questionnaires from their patients. There may be hundreds of questionnaire responses, but only four teams receiving the team-level intervention. Treating every response as independent evidence about the training programme could overstate the information available. A student should seek appropriate methodological advice before selecting an analysis, particularly where the number of groups is small.
Worked example: an area pattern is not a patient comparison
Imagine a fictional dissertation comparing neighbourhood deprivation scores with the proportion of residents receiving a preventive nursing contact. Each row represents a neighbourhood. The analysis could describe whether neighbourhoods with higher deprivation scores have different contact proportions. It could not establish that the most deprived individuals within each neighbourhood were the people who received, or missed, those contacts. Area averages do not reveal that individual relationship.
Suppose a neighbourhood has both a relatively high average deprivation score and high service uptake. Several explanations remain possible. Contacts might be concentrated among its better-resourced residents, access initiatives may have reached disadvantaged households, or population and recording differences may affect the figures. The area-level result alone cannot choose between those accounts. Your discussion should preserve the distinction and identify what linked individual information would be needed to investigate it.
A suitable conclusion would be: “In this fictional comparison, neighbourhood deprivation and recorded preventive contacts varied together at area level.” An unsuitable conclusion would say that deprived patients personally benefited more from the service. The latter moves from places to people without the necessary evidence. This error is commonly called the ecological fallacy; describing it accurately is more useful than merely adding its name to a limitations paragraph.
Worked example: individual interviews do not describe every service
Now consider interviews with twelve people recruited through one community clinic. These accounts might identify important experiences of appointment communication or continuity. They do not, by themselves, establish how every clinic in a region operates. Even a strong recurring theme across those interviews needs an explanation of whose accounts were included, how recruitment occurred and which contexts readers should consider when judging transferability.
The reverse mistake is sometimes called an atomistic fallacy: using individual-level observations to assume a relationship at a higher organisational or population level. For example, patients who describe helpful communication may also report greater confidence. That does not establish that clinics with higher average communication ratings will have better population outcomes. Organisational resources, case mix and other contextual factors could alter the relationship between services.
Repeated contacts and clustered observations need attention
Several records can come from the same person, and several people can receive care from the same team. Neither structure should disappear when a spreadsheet is assembled. If the question concerns patients, count distinct patients as well as contacts. If the question concerns service workload, contacts may be relevant, but describe repeat contacts and the period covered. A workload total and a patient coverage percentage answer different questions.
Clustering also matters when people share a household, practitioner, team or neighbourhood. Shared arrangements can make their observations more alike. Whether a particular analytical adjustment is suitable depends on the design, outcome, available groups and research question. Do not select a statistical method solely because its name sounds advanced. For a student project, a transparent descriptive analysis with clear limits can be more defensible than an unsupported model.
Build a level-of-analysis decision table
Before drafting the results, create a small planning table with the research question, row definition, denominator, intended comparison and permitted conclusion. Add a final column identifying what the data cannot show. This is a practical planning device, not a validated assessment scale. Its purpose is to expose mismatches early enough to revise the question or narrow the interpretation.
| Fictional question | Observation | Permitted focus | Important limit |
|---|---|---|---|
| How do patients describe access? | Individual interview | Participants’ experiences | Not the prevalence of experience across all services |
| How does team workload vary? | Team-period total | Recorded team workload | Not individual care quality |
| Where is service reach lower? | Area-period proportion | Geographical variation | Not individual causes of non-use |
Use consistent terminology across the proposal, methods, tables and conclusion. If “community” means a geographical area in the methods, do not let it become a claim about every person living there in the discussion. If “service” means one provider, name that boundary. These small wording decisions prevent the dissertation from suggesting a wider evidence base than it actually contains.
Keep the denominator and time window aligned
A proportion requires a numerator and denominator that refer to compatible populations and periods. For an area comparison, check whether residence, registration and service eligibility are defined in the same way. For a team comparison, check whether teams cover similar periods and whether reported activity includes the same kinds of contact. If definitions differ, document the discrepancy before interpreting the apparent pattern.
These checks do not require inventing a numerical correction. Sometimes the most honest approach is to report the available definitions, explain why direct comparison is limited and reduce the strength of the conclusion. An apparently precise percentage can be misleading when its underlying populations differ. Include the actual denominator alongside percentages so that readers can understand what each result represents.
Use nursing context without changing the analytical unit
A public health dissertation can remain clearly connected to nursing while analysing areas or organisations. Explain the relevant nursing responsibility, the service context and the decision that the analysis could inform. Then retain the correct unit in the results. A geographical pattern might justify asking where a nursing access review is needed; it does not prove that an individual nurse caused the pattern.
The separate question of how nursing activities might contribute to outcomes requires a contribution pathway. This article instead addresses the level at which observations and conclusions sit. Keeping those tasks distinct avoids two articles answering the same question. For broader framing, use the parent guide on keeping a public health dissertation focused on nursing.
A final supervisor discussion before analysis
Bring three items to your supervisor: one anonymised example of a data row, your proposed table headings and the exact conclusion you hope to write. Ask whether the observations support that conclusion at the stated level. If the answer is uncertain, revise the question before investing effort in a complex analysis. This discussion can identify mismatches that a polished introduction would otherwise conceal.
Finally, check the abstract independently. Abstracts often compress careful findings into stronger claims. Ensure that the population, setting, unit and time period remain visible even in the short summary. A reader should not need to reach the limitations section to discover that a claim about patients was based entirely on area-level information.
Check the unit in every results-table heading
Read each heading without its surrounding paragraph. Can a reader tell whether the figures concern people, contacts, teams or areas? Add the unit and period where needed. Then compare the heading with the first sentence of the discussion. If the subject changes from areas to patients, or from contacts to people, inspect the inference before submission. This small editorial check is especially useful when several datasets use similar labels for different units.
Keep a record of any unresolved mismatch and explain its effect on the conclusion. Clear wording cannot repair incompatible data, but it can prevent an avoidable overstatement.
Frequently asked questions
Is the unit of observation always the level of inference?
Not necessarily. A team intervention may generate patient observations, while the conclusion concerns teams. State the observation, intervention and inference levels separately. If they differ, explain the design implications and seek appropriate analytical advice.
What is an ecological fallacy?
It is an unsupported move from a group-level relationship to an individual-level conclusion. An area pattern does not show that the same relationship holds between people within that area. Keep the conclusion at the level supported by the observations.
What is the reverse error?
An individual-level relationship does not automatically describe relationships between organisations or areas. This is often termed an atomistic fallacy. Explain the actual mismatch rather than relying on a technical label without a worked example.
Do repeat contacts provide independent patient observations?
Not necessarily. Several records may concern the same person, and people may share a team or household. Identify repetition and clustering before analysis. The appropriate approach depends on the question, design and available groups.
What if the required data do not exist?
Consider narrowing the question or using a justified alternative source. Do not assume that collecting new interviews or patient information is automatically permitted. Any new data collection needs the relevant arrangements, and the revised question should match the evidence obtainable.
Review the analytical unit
Before submitting, read the abstract, results headings and conclusion together. They should refer to compatible units and make the same boundaries visible. A statement about an area should not become a statement about every resident simply because the summary is shorter. These fictional examples demonstrate analytical distinctions rather than findings from a real dataset. Show your supervisor the row definition and intended inference, particularly where repeated observations or clustered services are involved. Appropriate advice can help you decide whether to narrow the question or change the proposed analysis.
For the wider topic, read How to Keep a Public Health Dissertation Focused on Nursing. Explore the nursing dissertation resource hub for related guidance, or compare the available dissertation support services with your learning needs.
Students seeking help with community nursing research design can review our community and public health nursing dissertation support. Use the contact page to ask about educational feedback within your university’s rules. Support does not replace your own analysis, confer project approval or guarantee a grade.
References
Health Research Authority (2026) UK Policy Framework for Health and Social Care Research. Available at: Health Research Authority (Accessed: 16 October 2026).
