An example of evaluating a Housing Studies proposal research method design in terms of research methods design quality criteria

An example of evaluating a Housing Studies proposal research method design in terms of research methods design quality criteria (information from perplexity.ai)

My question: "Please comment on the research method design quality criteria (e.g. validity, reliability, dependability, etc.) of my Housing Studies research design:

To study the causes of domestic violence in the public housing estate in Tsuen Wan, Hong Kong:

To conduct narrative interviews on 5 domestic violence victims in Tsuen Wan public housing estate, who are my long-time public housing estate neighbors as well as to conduct semi-structured interviews on 1 social worker in Tsuen Wan on this topic. All the interviews are to be conducted in a Starbuck coffee shop in a Tsuen Wan shopping mall, for about 45 min. each."

The reply: "The proposed research design employs narrative and semi-structured interviews with a small, purposive sample of five domestic violence victims and one social worker, which aligns with qualitative exploratory aims but raises concerns across key quality criteria like credibility, transferability, dependability, and confirmability.​

Credibility

Credibility, akin to internal validity, is partially supported by using narrative interviews to capture victims' lived experiences and semi-structured ones for the social worker's insights, potentially providing rich, authentic data on causes in Tsuen Wan's public housing. However, interviewing long-time neighbors introduces significant researcher bias and power imbalances, as familiarity may inhibit candid disclosure on sensitive topics like domestic violence, undermining authenticity without member checking or triangulation.​

Transferability

Transferability, similar to external validity, is limited by the tiny sample size (n=6) and hyper-local focus on Tsuen Wan neighbors, making findings hard to apply beyond this specific context despite thick descriptions from narratives. No broader sampling strategy or demographic details further restrict generalizability to other Hong Kong public housing estates.​

Dependability

Dependability requires an audit trail, which is absent here—no mention of audio recording, verbatim transcription, reflexive journaling, or peer debriefing protocols. Conducting all 45-minute interviews in a public Starbucks risks distractions, incomplete data capture, and ethical breaches for traumatized participants, eroding consistency and repeatability.​

Confirmability

Confirmability demands neutrality, but researcher proximity as neighbors threatens objectivity, with no disclosed reflexivity on personal biases or positionality. Adding prolonged engagement, negative case analysis, or multiple data sources could strengthen this, but the design lacks them.​

Recommendations

To enhance quality, incorporate ethical safeguards like informed consent emphasizing anonymity, trauma-informed interviewing, and alternative private venues; expand to 10-15 diverse participants via snowballing; and detail thematic analysis with software like NVivo for rigor. These steps would better address validity gaps in qualitative domestic violence research."

A collection of blog notes on using chatgpt for research purpose.