Research in Social Work II, Qualitative and Mixed Methods: the Meaning and Tenets of Qualitative Research and its Difference from Quantitative Research, Designing Qualitative Research, its Methods from Field Study to Narratives, Managing and Analysing Qualitative Data, and the Components and Procedures of Mixed Methods Research

Designing a qualitative study involves choosing a topic and framing an open question ("How do families of children with autism experience the search for services?"); choosing an approach (ethnography, case study, phenomenology, grounded theory, narrative inquiry); selecting a site and participants by purposive, snowball or theoretical sampling; gaining entry through gatekeepers and building rapport; collecting data through in-depth interviews, observation, focus groups and documents; analysing data alongside collection; establishing trustworthiness; and writing. Sampling continues until saturation, the point at which new data no longer add new insights. Ethics are demanding because the relationship is close: informed and continuing consent, confidentiality when quotations might identify people, care with distressing topics, and awareness of the power gap between researcher and participant.

The methods the syllabus names are these. A field study, or ethnography, studies a group in its natural setting over an extended period through participant observation, informal interviews and documents; Bronisław Malinowski's fieldwork in the Trobriand Islands set the model, and William Foote Whyte's Street Corner Society (1943) applied it to an urban neighbourhood. A case study examines a bounded system, an individual, family, organisation, programme or community, in depth and in context, using multiple sources; Robert Stake distinguished intrinsic, instrumental and collective case studies, and Robert Yin stressed the case study for "how" and "why" questions about contemporary events. A focus group discussion brings together a small group, typically about six to ten participants with some shared characteristic, under a moderator to discuss a topic; its strength is the interaction, which reveals shared and contested views. Narrative methods, life histories, oral histories and biographical interviews, study the stories people tell about their lives and how they make sense of experience.

Observation may be participant or non-participant, overt or covert, structured or unstructured. Raymond Gold (1958) described four roles along a continuum: complete participant (fully involved, identity as researcher concealed), participant-as-observer (involved, with the research role known), observer-as-participant (mainly observing, with limited interaction) and complete observer (no interaction). Covert roles raise serious ethical problems. The syllabus's "Theoretic Research" is best read as research that builds theory from data. Its main form is grounded theory, introduced by Barney Glaser and Anselm Strauss in The Discovery of Grounded Theory (1967): rather than testing a theory, the researcher generates one grounded in data, through theoretical sampling, constant comparison and coding. Phenomenology, from Edmund Husserl, is a related approach that describes the essence of a lived experience, with the researcher bracketing (setting aside) their own assumptions.

Managing qualitative data begins with transcription of recordings, careful organisation and labelling of files, anonymisation, and field notes and memos that record observations and emerging ideas. Matthew Miles and Michael Huberman (1994) described analysis as three concurrent flows of activity: data reduction (selecting, focusing and coding), data display (matrices, networks and charts) and conclusion drawing and verification. Coding assigns labels to segments of data. In Anselm Strauss and Juliet Corbin's grounded theory (1990), open coding breaks data into concepts, axial coding relates categories to each other around a central category, and selective coding integrates them into a theory; the constant comparative method compares each new piece of data with those already coded. Virginia Braun and Victoria Clarke's thematic analysis (2006) sets out six phases: familiarisation with the data, generating initial codes, searching for themes, reviewing themes, defining and naming themes, and producing the report. Other techniques include qualitative content analysis, narrative analysis, discourse analysis and framework analysis. Software such as NVivo, ATLAS.ti and MAXQDA helps organise and retrieve coded data, but does not do the interpretation.

Yvonna Lincoln and Egon Guba (Naturalistic Inquiry, 1985) proposed trustworthiness as the qualitative counterpart of validity and reliability, with four criteria. Credibility (parallel to internal validity) is established by prolonged engagement, persistent observation, triangulation, peer debriefing and member checking, returning findings to participants. Transferability (parallel to external validity) is supported by thick description that lets readers judge whether findings apply elsewhere. Dependability (parallel to reliability) is shown by an audit trail of decisions. Confirmability (parallel to objectivity) is supported by reflexivity and the audit trail. Norman Denzin distinguished four types of triangulation: data triangulation (different sources, times or places), investigator triangulation (several researchers), theory triangulation (several theoretical perspectives) and methodological triangulation (several methods). The qualitative report presents themes supported by participants' own words, describes the context richly, explains the researcher's position and methods openly, and protects participants' identities.

Mixed methods research collects and analyses both quantitative and qualitative data in one study and integrates them, on the argument that each compensates for the other's weaknesses: numbers give breadth and generalisation, words give depth and meaning. Its philosophical base is usually pragmatism, which judges methods by what answers the question. John Creswell and Vicki Plano Clark describe its components as the priority or weighting given to each strand, the timing or sequence of the strands (concurrent or sequential), and the point and manner of mixing or integration (in design, data collection, analysis or interpretation). Janice Morse's notation writes the dominant strand in capitals, "+" for concurrent and "→" for sequential: QUAN → qual, for instance, is a sequential design led by the quantitative strand.

The procedures of combining the two strands include connecting (using the results of one to design the other, as in sequential designs), merging (bringing the two datasets together in analysis, for example in a joint display that sets statistics beside themes), and embedding (a supporting strand within a larger design, such as interviews within a trial). Integration is the step most often neglected: a study that reports a survey and interviews side by side without relating them is two studies, not a mixed methods one. Mixed methods suit programme evaluation especially well, where both whether a programme works and how and for whom it works matter.