Understanding Sample Sizes in Convergent Mixed Methods Research

In a convergent (or parallel) mixed methods design, qualitative and quantitative data are collected simultaneously and analyzed separately before integration (e.g., merging findings to explain or corroborate each other). Sample sizes aren't one-size-fits-all—they depend on your research questions, population, resources, feasibility, and the need for saturation (in qualitative) or statistical power (in quantitative). There's no universal formula, but guidelines from experts like Creswell and Plano Clark (2018) or Fetters et al. (2013) emphasize balancing depth and breadth while ensuring the samples support integration.

I'll break this down by strand, providing evidence-based recommendations. For precise planning, consider consulting a statistician or using tools like G*Power for quantitative power analysis.

1. Quantitative Strand Sample Size

The quantitative component typically aims for generalizability, reliability, and statistical power to detect effects. Sample size is driven by:

• Statistical power: Usually target 80% power to detect a medium effect size (e.g., Cohen's d = 0.5) at α = 0.05.


• Study type: Surveys, experiments, or correlational analyses require different sizes.


• Expected response rate and variability: Account for attrition (e.g., 20-30% in surveys).

Guidelines:

• Small-scale studies (e.g., pilot or exploratory): 100-300 participants for basic descriptive stats or simple regressions.


• Medium-scale (e.g., testing hypotheses with t-tests or ANOVA): 200-500, depending on subgroups (e.g., add 30-50 per variable or group).


• Large-scale (e.g., complex models like multiple regression or SEM): 500+; aim for 10-20 participants per variable/predictor.


• Evidence: Based on Cohen's (1988) power tables and modern simulations. For example, a two-group t-test needs ~64 per group for medium effect; surveys often require n=385 for a ±5% margin of error at 95% confidence (assuming infinite population).

Practical Steps:

• Use power analysis software (e.g., G*Power) with your expected effect size from prior studies.


• If your population is small (e.g., <500), use finite population corrections to avoid oversampling.


• In mixed methods, ensure the quantitative sample is large enough to provide robust benchmarks for qualitative insights—e.g., if qualitative explores subgroups, quantitative should cover them proportionally.

2. Qualitative Strand Sample Size

Qualitative data focuses on depth and saturation (when no new themes emerge). In convergent designs, it's often smaller to allow simultaneous collection without overwhelming logistics. Size depends on:

• Data collection method: Interviews, focus groups, or observations.


• Homogeneity of participants: More homogeneous groups need smaller samples.


• Complexity: Richer topics (e.g., sensitive health experiences) may require larger samples for saturation.

Guidelines:

• Interviews: 5-25 participants per subgroup (e.g., 10-15 total for broad themes; up to 25 for diverse perspectives). Saturation often occurs around 12-17 for semi-structured interviews.


• Focus groups: 3-6 groups, each with 6-10 participants (total ~20-50 people).


• Observations or documents: 10-30 units, guided by thematic saturation rather than a fixed number.


• Evidence: Guest et al. (2006) found saturation in interviews typically by 9-17 participants (mean 12) for most studies. Hennink et al. (2017) recommend monitoring for diminishing returns after 20-30. In mixed methods, aim for enough to illuminate quantitative patterns—e.g., 15-20 interviews to explain survey trends.

Practical Steps:

• Start with 8-12 and add until saturation (track via thematic analysis software like NVivo).


• In convergent designs, align qualitative sampling with quantitative (e.g., purposive sampling from the quant pool for deeper dives).


• Ensure diversity (e.g., by demographics) to support integration, but avoid over-sampling to maintain feasibility alongside quant data collection.

Integrating Sample Sizes in Your Study

• Total sample: Often, the quantitative sample is larger (n=200-1000+), with a subset (10-20%) selected for qualitative (e.g., 200 quant respondents, 20 for interviews). This enables concurrent collection and easy merging.


• Feasibility check: Consider budget, time, and ethics (e.g., IRB limits). Pilot test to refine.


• Common pitfalls: Underpowering quantitative (leads to null results) or undersampling qualitative (misses nuances). If your study involves underrepresented groups, prioritize ethical inclusivity over strict sizes.


• Resources for tailoring: Refer to Tashakkori and Teddlie (2010) for mixed methods specifics, or online calculators like Qualtrics sample size tool for quant.

If you provide more details (e.g., your research topic, methods, or population size), I can refine these recommendations further.