A typical mixed methods thesis: Chapter 4 reports a survey of 350 teachers, Chapter 5 reports interviews with 15 teachers, and Chapter 6 discusses each chapter in turn. By the end, the reader cannot tell what the interviews added to the survey numbers, or whether the numbers confirmed or contradicted the stories. Those are two studies stapled together.
The value of mixed methods lies in integration: the point where the two kinds of data are brought together to produce insight that neither could deliver alone. Without it you pay double the effort and gain nothing extra.
When you genuinely need mixed methods
- Quantitative results need explaining: why one group differs from another, why there are outliers.
- No suitable instrument exists for your context, so qualitative exploration must come first to build one.
- You want to compare two perspectives on the same phenomenon to see whether they converge or diverge.
- You are evaluating a programme and need both its effects and an understanding of how it was implemented.
If one type of data fully answers your question, do not add a second just to make the proposal look richer.
Three core designs
| Design | Sequence | Purpose | Point of integration |
|---|---|---|---|
| Convergent | Quantitative and qualitative in parallel | Compare and corroborate two sources | At analysis: results placed side by side |
| Explanatory sequential | Quantitative first, qualitative second | Explain quantitative results | When sampling and drafting interview questions from survey results |
| Exploratory sequential | Qualitative first, quantitative second | Build an instrument or test generalisability | When turning qualitative themes into items and variables |
The explanatory sequential design is the most practical for a thesis because each phase has a clear output and phase two is guided by phase one. The exploratory sequential design is the most time-consuming because building a new instrument requires extra rounds of testing.
Writing a mixed methods research question
Besides separate quantitative and qualitative questions, include an integrative question that can only be answered by combining both kinds of data. For example:
- Quantitative: How does classroom technology use vary with years of teaching experience?
- Qualitative: What do teachers describe as shaping their decisions to use technology?
- Integrative: How do teachers’ experiences help explain the experience-related differences observed in the survey?
An integrative question forces you to plan the connection from the outset rather than hoping it appears in the discussion chapter.
Linked sampling across phases
In an explanatory sequential design, interviewees should be selected on the basis of survey results. For instance, choose four people whose technology use is unusually high for their experience band, four unusually low, and four typical. To make this possible, the questionnaire needs a consent-to-recontact item, with contact details stored separately from responses to protect confidentiality.
Selecting interviewees at random or by convenience, without reference to phase one, throws away the single most important integration point of the design.
The joint display
The most useful integration tool is the joint display: each row is a finding or subgroup, and the columns place quantitative results, qualitative results and a meta-inference side by side. One row might look like this:
| Group | Survey result | Interview result | Meta-inference |
|---|---|---|---|
| Teachers with 10–15 years’ experience | Lowest technology-use scores of all groups | Describe established lesson systems and reluctance to redesign without preparation time | Low use reflects switching costs more than skill gaps; support should target time, not training |
The last column is something only mixed methods can produce. If it is empty, or merely restates the other two, you have not integrated.
When the two strands disagree
The survey shows students are highly satisfied with lecturer feedback; interviews reveal they rarely read it. That is not a failure but a finding. Perhaps the survey item captured satisfaction with the lecturer’s attitude while the interviews addressed usefulness. Report the divergence, propose an explanation and, where possible, return to the data to test it.
Common pitfalls and how to avoid them
- Decorative qualitative data: a few interview quotes illustrating the numbers without full analysis.
- Unnamed design: readers cannot tell the sequence, priority or integration point.
- Underestimating time: sequential designs require phase one analysis to finish before phase two begins, and your timeline must allow for that.
- A split discussion: discuss by research question, not by data type.
A checklist for a mixed methods proposal
- The reason for mixing, stated in one sentence.
- The named design and a sequence diagram.
- Three questions: quantitative, qualitative and integrative.
- Specific integration points: sampling, instrument building or analysis.
- A planned joint display.
- A timeline with an analysis gap between phases.
Next step: sketch an empty joint display for your project now, with rows for the findings you expect. If you cannot imagine what the meta-inference column would contain, reconsider whether your project really needs mixed methods.
Câu hỏi thường gặp
What is mixed methods research?
It is a design that combines quantitative and qualitative data in one study with a deliberate integration step. Presenting both kinds of data side by side without connecting them is not mixed methods in the full sense.
What is an explanatory sequential design?
It collects and analyses quantitative data first, then uses qualitative data to explain the results. Phase two participants are usually selected on the basis of phase one findings.
What is a joint display in mixed methods?
It is a table or figure that places quantitative results, qualitative results and integrated meta-inferences side by side for each finding or group. It is a widely used way to show integration.
What if my quantitative and qualitative results contradict each other?
Report the divergence as a finding, propose explanations such as differences in what each method captured, and check them against the data where possible. Do not hide it or pick only one source.
Is mixed methods a good choice for a master’s dissertation?
Only if the question truly needs both kinds of data and the timeline allows sequential analysis. A small-scale explanatory sequential design is usually the most feasible option.
