How do I integrate qualitative and quantitative data in a mixed methods study?

The failure mode is easy to describe and hard to notice from inside: a thesis with a survey chapter and an interview chapter, both competent, that never speak to each other. That is two studies sharing a binding, and reviewers say so.

Integration has to be designed rather than added later. Decide at the planning stage what each strand is for and where they will meet. In an explanatory design the meeting point is the interview guide, which should be built from the quantitative results, including the results that surprised you. In an exploratory design the meeting point is the instrument, whose items come from the qualitative categories.

Methodologists call that meeting point the point of integration, or point of interface, and they separate a results point of integration from an analytical one (Schoonenboom and Johnson, 2017, p. 115). Four connection strategies are available to you: merge the two datasets, use the analysis of one strand to direct collection in the other, embed one form of data inside a larger design, or bind the datasets with a theoretical or program framework (Schoonenboom and Johnson, 2017, p. 116). Pick the strategy before you collect anything, because two of the four constrain what you can collect later.

The joint display is the most practical tool. A table with one row per finding, a column for the quantitative evidence, a column for the qualitative evidence, and a column for your interpretation of the two together. Filling it in forces the integration to happen, and the finished table usually earns a place in the thesis. Make the third column an explicit integrative statement saying whether the strands converge, complement one another, or diverge, rather than a summary of the first two (Schoonenboom and Johnson, 2017, pp. 122-123). You are also not limited to one meeting point. The number and location of integration points should be constructed to fit your research questions (Schoonenboom and Johnson, 2017, p. 115).

Pay particular attention to disagreement between strands. When the survey says one thing and the interviews say another, that divergence is the most interesting result in the study, and explaining it is usually where the contribution lies.

A published example shows how far divergence can go. In McMahon's study of rape myths among college student athletes, a survey of 205 sophomores and juniors at one northeastern public university found very low acceptance of rape myths, with higher acceptance of violence among men and among participants who did not know a survivor of sexual assault. The follow-up focus groups and interviews produced four themes that contradicted the survey, including misunderstanding of consent, belief in accidental or fabricated rape, and claims that some women provoke rape (Schoonenboom and Johnson, 2017, pp. 126-127). Because the qualitative strand contradicted rather than explained the quantitative one, the label "sequential explanatory" was inaccurate for that study, and "sequential-comparative" or "sequential-triangulation" described it better. Before you try to resolve divergence of that kind, check that it is genuine, audit the inferential strength of each strand separately, and only then reach for further research, broader theorizing, or reanalysis (Schoonenboom and Johnson, 2017, p. 116).

What are the standard mixed methods designs?

Three, distinguished by sequence and purpose. Convergent, where both strands run in parallel and are compared. Explanatory sequential, where quantitative results are followed by qualitative work explaining them. And exploratory sequential, where qualitative work builds an instrument or hypothesis the quantitative strand then tests.

Name your design and cite a methodological source for it. Examiners in mixed methods work expect the vocabulary, and using it correctly signals that the combination was a decision rather than an accumulation.

State the priority between strands too, since one usually leads. A design where the qualitative strand is subordinate is legitimate, and pretending both were equal when they were not creates a mismatch a reader can see.

The three-design vocabulary is standard, but it is not the full list. A conversion design is a fourth option, where you transform qualitative categories or themes into quantitative variables and analyze them statistically, which puts the integration inside the analysis rather than after it (Schoonenboom and Johnson, 2017, p. 114). If your data suits that, say so, because forcing it into the convergent or sequential labels will misdescribe what you actually did.

Where does integration appear in the write-up?

In a dedicated section, usually after both strands are reported, and again in the discussion. Naming it as an integration section signals to a reader that you did the work. Leaving integration implicit is the most common reason a study gets called multi-method rather than mixed methods.

The discussion then works at the level of the combined finding rather than strand by strand. If you find yourself writing "the survey showed X" and "the interviews showed Y" in separate paragraphs there, the integration section did not do its job.

Report the points where integration was attempted and failed, if there were any. A qualitative theme with no quantitative counterpart is a real result about the limits of your instrument, and saying so is more useful than leaving it out.