4.4 Sampling and integration challenges
As discussed in the soup analogy, quantitative and qualitative approaches rely on different sampling logic. Quantitative research typically uses random sampling to find a representative ‘spoonful’, while qualitative research hunts for specific ingredients using purposive sampling to understand textures and flavours. In mixed methods you must align your sampling strategy with the specific strand of your design.
Activity 4.3: Making sense of ‘high achievers’
Read the following research scenario:
You are investigating student engagement in a fully online undergraduate course where ‘engagement’ is being treated initially as time-on-platform and click-through behaviour.
- QUAN data: Learning analytics from 500 online students show that the ‘high achievers’ (students with top grades) actually have the lowest ‘time spent on page’ and the fewest clicks.
- QUAL data: Follow-up interviews with ten high achievers reveal that they find the online platform ‘cluttered and distracting’ so they download the PDFs of the course content and study offline.
Consider the following questions:
- What type of mixed methods design is this?
- How would you select the 500 students in the QUAN phase and the ten students in the QUAL phase?
- What should you do when findings seem contradictory? Do you trust the QUAN (N=500) or the QUAL (N=10)?
- How do you integrate these findings to redefine what ‘online engagement’ actually looks like and at what stage(s) of the study does that integration occur?
Answer
- This is an explanatory sequential design (QUAN → qual). The quantitative data (learning analytics) identified a strange pattern and the qualitative interviews followed up to explain it.
- The students would be selected as follows:
- 500 students (quantitative strand): To select the sample of 500 from the wider population of online students you would use random sampling. In the soup analogy this is like wanting to know if a pot of soup is too salty: you stir the pot well and take a small, random spoonful. Because it is random, that spoonful represents the whole pot, allowing you to safely generalise the learning analytics (like clicks and time spent) to the broader student population.
- Ten students (qualitative strand): To select the ten high achievers to interview you would use purposive sampling. Returning to the soup analogy, you are no longer checking the overall saltiness; you want to know what specific ingredients are in the soup. So you specifically choose these ten students because they possess the unique ‘textures and flavours’ (the contradictory behaviour of high grades but low page time) that will help you to understand this specific phenomenon in depth.
- Both findings are ‘true’. You trust the QUAN data to tell you what is happening (low clicks) and the QUAL data to tell you why (studying offline).
- You integrate these findings by redefining ‘engagement’. The study shows that digital footprint (clicks) is a flawed proxy for actual cognitive engagement. High achievers are highly engaged with the material but not with the online platform hosting it. Integration in this study happens at two stages: first at the design stage (the QUAL phase is set up specifically to explore the QUAN pattern), and again at the interpretation stage where the two strands are brought together to redefine the construct.
4.3 Frameworks that can help ensure rigour


