Designing Mixed Methods Research

A mixed methods study needs three decisions made explicit before data collection begins:

1. Purpose for mixing — why can't one method alone answer the question? (e.g., triangulation, complementarity, expansion, development)
2. Priority & sequence — is QUAL or QUANT dominant, and does one phase come before the other, or do they run concurrently?
3. Point of integration — where do the two strands actually meet (design, methods, or interpretation/results)?

Example one-line design statement:

"This is an explanatory sequential design (QUANT → qual) where survey results on caseworker burnout inform the sampling and interview guide for a follow-up qualitative phase, integrated through a joint display at the interpretation stage."

Progress:

• Step 1: Clarify the research question and confirm it truly requires both data types
• Step 2: Select a design (convergent, explanatory sequential, exploratory sequential, or embedded/multiphase)
• Step 3: Determine priority (QUAL+quant, QUANT+qual, or equal weight) and timing (concurrent vs. sequential)
• Step 4: Plan the integration point and integration technique
• Step 5: Address sampling, rigor, and ethics for each strand
• Step 6: Collect and analyze each strand per its own methodological standards
• Step 7: Integrate — merge, connect, or embed findings
• Step 8: Report using mixed methods standards (e.g., APA JARS-Mixed) with a joint display or meta-inference

Step 1: Confirm the need for mixing

Ask: what would qualitative data reveal that quantitative can't (and vice versa)? Common justifications:

• Triangulation — corroborate findings across methods
• Complementarity — elaborate/clarify results from one method with another
• Development — use one method's results to build the other (e.g., qual informs survey items)
• Expansion — extend breadth/range of inquiry by using different methods for different components
• Initiation/divergence — surface contradictions that prompt reanalysis

If none apply, a single-method design is likely more appropriate — don't mix methods just to seem rigorous.

Step 2: Select a design

• Convergent (parallel): collect qual and quant simultaneously, independently analyze, then merge/compare at interpretation. Best for triangulation.
• Explanatory sequential (QUANT → qual): quantitative results first, then qualitative data explains unexpected or significant findings.
• Exploratory sequential (QUAL → quant): qualitative exploration first (e.g., to build a construct or instrument), then quantitative testing/generalization.
• Embedded/multiphase: one method nested within a larger design (e.g., qual interviews embedded within an RCT to explain implementation) or multiple phases across a long-term project/evaluation.

Step 3: Set priority and timing

State explicitly which strand is dominant (if any) using notation: capital letters = priority (QUAL/QUANT), lowercase = supplemental (qual/quant), arrows = sequence, plus sign = concurrence. E.g., QUAL → quant or QUAN + QUAL.

Step 4: Plan integration

Integration is the hallmark of real mixed methods work — parallel reporting of two studies is NOT mixed methods. Techniques:

• Merging: side-by-side comparison, joint displays, data transformation (quantitizing/qualitizing)
• Connecting: one strand's results directly inform the sampling, protocol, or instrument of the next
• Embedding: one strand supports the other within a single, larger design (e.g., qual data explaining outliers in a quant model)

Use a joint display (table or graphic) to show quant results next to qual quotes/themes and draw a meta-inference — the integrated conclusion that neither strand could produce alone.

Step 5: Sampling, rigor, ethics

• Sampling can differ by strand (e.g., probability sample for survey, purposive subsample for interviews) — justify the relationship between the two samples (identical, nested, parallel, or multilevel).
• Apply quantitative rigor standards (validity, reliability) to the quant strand and qualitative rigor standards (credibility, trustworthiness, reflexivity) to the qual strand — don't force one paradigm's criteria onto the other.
• Address informed consent and confidentiality for both data types; qualitative data (interviews, open-text) often carries higher re-identification risk and needs tailored protections.
• Acknowledge the researcher's paradigm stance (pragmatism is the most common philosophical home for mixed methods, treating qual/quant as complementary tools rather than competing worldviews).

Step 6–7: Analyze and integrate

• Analyze each strand independently first, using appropriate software/techniques (e.g., statistical software for quant; NVivo/MAXQDA/ATLAS.ti for qual, several of which have built-in mixed methods merging tools).
• Then bring results together at the planned integration point; look explicitly for confirmation, expansion, or discordance between strands. Discordant results are valuable, not a flaw — they should be reported and interpreted, not hidden.

Step 8: Report

Use recognized reporting standards (e.g., APA JARS-Mixed) and include:

• Explicit statement of design type, priority, and timing
• Rationale for mixing
• Description of integration procedures and a joint display
• Meta-inferences drawn from the integrated data
• Limitations specific to mixing (e.g., unequal sample sizes, timing constraints, integration challenges)

Example 1: Input: "We ran a client satisfaction survey (n=200) and want to also understand why dissatisfied clients felt that way." Output: Explanatory sequential design (QUANT → qual). Phase 1: analyze survey for satisfaction predictors/outliers. Phase 2: purposively sample low-satisfaction respondents for semi-structured interviews. Integration point: interpretation, via joint display connecting survey predictors to interview themes explaining the "why" behind low scores.

Example 2: Input: "We don't have a validated scale for 'caseworker moral distress' — how do we build one?" Output: Exploratory sequential design (QUAL → quant). Phase 1: conduct focus groups/interviews to identify dimensions of moral distress. Phase 2: use themes to draft survey items, pilot test, then validate with a larger quantitative sample (factor analysis). Integration point: instrument development (connecting), plus a later validation report showing how qual themes map onto quant factors.

Example 3: Input: A manuscript reports a survey in the results section and separately reports interview themes in a different section, with no comparison. Output: This is NOT integrated mixed methods — it is two parallel single-method studies. Recommend adding a joint display comparing quant and qual results side-by-side and writing explicit meta-inferences that state what the combination reveals beyond either method alone.