Sample PhD Methodology Chapter: A Mixed-Methods Research Design

Sample PhD Methodology Chapter: A Mixed-Methods Design for Investigating Employee Wellbeing in Remote Work

3.1 Introduction to the Chapter

The preceding chapters established the conceptual foundations of this thesis. Chapter 1 framed the problem of employee wellbeing in the context of the accelerated and, for many organisations, permanent shift towards remote and hybrid working arrangements. Chapter 2 developed a critical review of the literature, identifying that while a substantial body of work links remote work to outcomes such as autonomy, work intensification and social isolation, the mechanisms through which these forces combine to shape subjective wellbeing remain underspecified, and are rarely examined with designs that integrate breadth of measurement with depth of interpretation. The overarching aim of the research is therefore to explain how, and under what boundary conditions, remote working practices influence the psychological, social and physical dimensions of employee wellbeing.

This chapter sets out and justifies the methodological architecture through which that aim is pursued. Its purpose is not merely to describe a set of procedures but to demonstrate that each procedural choice follows coherently from a defensible position on the nature of knowledge and the requirements of the research questions. At doctoral level the methodology chapter carries a distinctive evidentiary burden: it must render the entire inquiry transparent, auditable and defensible against the examiner’s central question of whether the conclusions can reasonably be trusted. Following the logic of the “research onion” advanced by Saunders, Lewis and Thornhill (2019), the chapter proceeds from the outer, more abstract layers of philosophy and approach inward towards the concrete choices of instrument, procedure and analytical technique, so that the reader can trace an unbroken line from epistemological commitment to analytical output.

The research is guided by three questions. First, to what extent do specific features of remote work — spatial autonomy, temporal flexibility, communication load and boundary permeability — predict variation in employee wellbeing? Second, through what experienced processes do employees themselves account for the effect of remote work on their wellbeing? Third, how do the quantitative patterns and the qualitative accounts converge, diverge or elaborate one another to produce an integrated explanation? The first question is broadly explanatory and variance-oriented; the second is interpretive and process-oriented; the third is integrative. This heterogeneity of questions is itself the first signal that a single-method design would be insufficient, and it foreshadows the mixed-methods logic developed below.

The chapter is organised as follows. Section 3.2 articulates and defends the research philosophy of pragmatism. Section 3.3 sets out the mixed-methods approach and the sequential explanatory design. Sections 3.4 and 3.5 detail the population, sampling strategy and data-collection instruments for both strands. Section 3.6 reports the pilot study. Section 3.7 explains the analytical procedures. Section 3.8 addresses validity, reliability and trustworthiness; Section 3.9 sets out the ethical framework; and Section 3.10 acknowledges the methodological limitations before the chapter is summarised in Section 3.11.

3.2 Research Philosophy

Every research design rests, whether acknowledged or not, on assumptions about the nature of reality (ontology) and about what can be known and how (epistemology). Making these assumptions explicit is a condition of methodological rigour, because they determine what will count as legitimate evidence and as valid inference (Crotty, 1998; Saunders, Lewis and Thornhill, 2019). This study is located within the philosophy of pragmatism. The remainder of this section justifies that position by contrasting it with the two paradigms it must displace: positivism and interpretivism.

Positivism holds that social reality exists independently of the observer and can be measured objectively through procedures modelled on the natural sciences; the researcher stands apart from the object of study, values are bracketed out, and the aim is the discovery of law-like regularities through hypothesis testing and statistical generalisation (Bryman and Bell, 2015). A purely positivist design applied to the present research would have considerable strengths: it would permit the measurement of wellbeing across a large sample, the estimation of relationships between remote-work features and wellbeing outcomes, and generalisation to a wider population. Yet it would also be limited in a way that is fatal to the second research question. Positivism struggles to access the meanings employees attach to their experience — the felt sense of isolation, the negotiated boundary between home and work, the way autonomy is experienced as either liberating or abandoning depending on context. These meanings are not epiphenomenal to the research problem; they are central to it, because wellbeing is in part a subjective, interpreted state.

Interpretivism occupies the opposing pole. It treats social reality as constructed through the meanings that actors bring to their situations, holds that these meanings can only be accessed through empathetic, contextualised understanding, and rejects the notion of a value-free observer (Bryman and Bell, 2015; Crotty, 1998). An interpretivist design would be well suited to the second research question, generating rich accounts of how employees make sense of remote work. But interpretivism, taken alone, cannot answer the first question. It offers no basis for estimating the extent to which particular features of remote work are associated with wellbeing across a population, nor for assessing whether patterns observed in a handful of accounts hold more widely. Committing exclusively to interpretivism would sacrifice the explanatory generalisation the thesis requires.

The research questions therefore pull in two directions at once, and neither classical paradigm can accommodate both without distortion. Pragmatism resolves this tension. Rather than beginning from a fixed ontological commitment, pragmatism begins from the research problem and treats the question of which methods to use as one to be settled by what will most usefully answer it (Tashakkori and Teddlie, 1998; Morgan, 2007). Its intellectual lineage runs through the American pragmatist tradition — Peirce, James and, most influentially for methodology, Dewey — for whom the meaning and value of an idea lie in its practical consequences (Morgan, 2007). Ontologically, pragmatism is neither strictly realist nor strictly relativist; it accepts that there is a real world that constrains inquiry while holding that our knowledge of it is always mediated, provisional and shaped by the purposes for which it is sought. Epistemologically, it rejects the forced choice between objective measurement and subjective understanding, treating both numerical and narrative evidence as legitimate provided they are gathered rigorously and serve the inquiry.

Three features make pragmatism the appropriate philosophy for this study. First, it dissolves the incompatibility thesis — the claim that quantitative and qualitative methods rest on irreconcilable paradigms and cannot be combined — which would otherwise prohibit the mixed design the research questions demand (Tashakkori and Teddlie, 1998; Johnson and Onwuegbuzie, 2004). Second, its problem-centred orientation matches the applied ambition of the thesis, which seeks not only to explain wellbeing but to inform the design of humane remote-work policy; pragmatism’s criterion of “what works” aligns knowledge production with practical utility (Feilzer, 2010). Third, it provides a principled rather than merely opportunistic justification for methodological pluralism: methods are chosen because they answer the questions, and the resulting knowledge claims are evaluated by their coherence, credibility and usefulness rather than by fidelity to a single paradigm. For these reasons the study adopts pragmatism as its guiding philosophy, and it is from this position that the mixed-methods approach set out below follows.

3.3 Research Approach and Design

Consistent with its pragmatist foundation, the study adopts a mixed-methods approach, defined as research in which the investigator collects and analyses both quantitative and qualitative data, integrates the two, and draws inferences using the combined strengths of each to understand the problem (Creswell and Plano Clark, 2018). Mixed methods is not simply the co-presence of numbers and words; its defining feature is integration, the deliberate bringing together of the two strands so that the combined account is more complete than either alone (Bryman, 2006). The rationale here is principally one of complementarity and expansion: the quantitative strand establishes the extent and shape of relationships between remote-work features and wellbeing across a population, while the qualitative strand explains the processes and meanings that give those relationships their content, extending the breadth and range of the inquiry (Greene, Caracelli and Graham, 1989).

The approach to theory is best characterised as abductive rather than purely deductive or inductive. The quantitative strand is broadly deductive, deriving testable hypotheses from the job demands–resources framework and related wellbeing theory reviewed in Chapter 2. The qualitative strand is broadly inductive, remaining open to meanings and processes not anticipated by that framework. Abduction, which moves iteratively between theoretical expectation and empirical surprise to arrive at the most plausible explanation, captures the overall logic and is well suited to pragmatist mixed-methods work (Morgan, 2007; Saunders, Lewis and Thornhill, 2019).

Within the family of mixed-methods designs catalogued by Creswell and Plano Clark (2018), the study employs a sequential explanatory design (also termed the explanatory sequential design). This design unfolds in two connected phases. In the first phase, quantitative data are collected and analysed to establish the pattern of relationships across the population. In the second phase, qualitative data are collected and analysed specifically to explain, elaborate and add depth to the quantitative results — in particular, to illuminate findings that are statistically notable, unexpected or ambiguous. The qualitative phase is thus not free-standing but is designed to build directly on the quantitative results, with the connection between phases occurring at the point of participant selection: cases for interview are chosen on the basis of their standing on the quantitative measures.

The sequential explanatory design was selected over the plausible alternatives for reasons that follow directly from the research questions. A convergent parallel design, in which both strands are collected concurrently and then merged, would have answered the first two questions but would have forfeited the central analytical advantage the thesis seeks: the ability to use qualitative data to explain specifically those quantitative results that most require explanation. Because interview participants are sampled on the basis of survey responses, concurrent collection is logically impossible — the second phase depends on the output of the first. An exploratory sequential design, which reverses the order by beginning qualitatively (typically to build an instrument), was rejected because validated wellbeing instruments already exist and the study’s priority is explanation of measured relationships rather than instrument development. The explanatory sequence therefore uniquely fits a study whose first task is to map relationships and whose second is to explain them.

The design does carry a well-recognised cost. Because the phases are sequential rather than parallel, the overall study is more time-consuming, and the researcher cannot finalise the qualitative sampling frame or interview foci until the quantitative analysis is complete (Creswell and Plano Clark, 2018). This is accepted as a reasonable price for the interpretive leverage the design provides, and the fieldwork timetable in Section 3.4 is structured to accommodate it. In terms of priority and weighting, the study gives the two strands roughly equal status, with the quantitative strand establishing the terrain and the qualitative strand carrying much of the explanatory burden; the point of interface is the selection of interviewees, and integration is completed at the interpretation stage through joint displays and a meta-inference, as detailed in Section 3.7.

3.4 Population and Sampling

3.4.1 Target Population and Setting

The target population comprises knowledge workers employed in organisations that operate remote or hybrid working arrangements. Knowledge work is the appropriate focus because it is the segment of the labour market in which remote work is both feasible and prevalent, and in which the boundary and autonomy dynamics central to the research questions are most salient. To bound the study and control for gross contextual heterogeneity, the accessible population is defined as employees of medium-to-large organisations in the professional, financial and technology services sectors, who have worked remotely for at least one full day per week over the preceding six months. This tenure criterion ensures that participants have sufficient experience of remote working to report on it meaningfully.

3.4.2 Quantitative Sampling

The quantitative phase aims for a probability sample where feasible, and a large, demographically diverse non-probability sample where organisational access constrains random selection. In the primary access route, participating organisations provide a sampling frame of eligible employees from which a stratified random sample is drawn, with strata defined by remote-work intensity (fully remote, hybrid) and by job family, so that the sample reflects the composition of the population on variables theoretically related to wellbeing (Bryman and Bell, 2015). Where a complete frame cannot be obtained, the study supplements this with an organisationally endorsed census invitation to all eligible employees, which, while not a random sample, reduces self-selection relative to open recruitment.

Sample size is determined a priori through statistical power analysis rather than by rule of thumb. For the planned multiple regression and structural analyses, an a priori power calculation was conducted for a conventional target of statistical power of 0.80 at an alpha level of 0.05, seeking to detect a small-to-medium effect. This calculation, together with an allowance for a proportion of incomplete responses and for the multivariate techniques planned, indicates a minimum usable sample in the region of 300–380 complete responses (Field, 2018; Tabachnick and Fidell, 2019). Recruiting to the upper end of this range provides a margin for listwise deletion and permits the subgroup comparisons the analysis requires.

3.4.3 Qualitative Sampling

Because the second phase is designed to explain the first, its participants are drawn from survey respondents using purposive, criterion-based sampling guided by the quantitative results (Creswell and Plano Clark, 2018; Palinkas et al., 2015). At the close of the survey, respondents indicate whether they are willing to be contacted for a follow-up interview and provide contact details separately from their questionnaire responses. From this consenting pool, the researcher selects a maximum-variation subsample that deliberately spans the range of wellbeing scores — including respondents with high and low wellbeing, and cases whose scores are unexpected given their remote-work profile (for example, high autonomy but low wellbeing). Sampling for informative contrast in this way maximises the explanatory yield of the interviews.

The qualitative sample size is governed by the principle of information power (Malterud, Siersma and Guassora, 2016) and by the goal of data saturation, the point at which additional interviews cease to generate new codes or themes (Guest, Bunce and Johnson, 2006). On the basis of the narrow aim, the reasonably specific sample and the theory-informed interview guide, a sample of approximately 20–25 interviews is anticipated to be sufficient, with recruitment continuing until saturation is judged to have been reached and documented rather than assumed. The precise number is therefore treated as an outcome of the fieldwork, not a fixed target set in advance.

3.5 Data Collection Instruments

3.5.1 Quantitative Instrument: The Survey Questionnaire

Quantitative data are gathered through a structured, self-administered online questionnaire. Wherever possible the questionnaire uses established, previously validated multi-item scales rather than bespoke items, because validated instruments carry accumulated evidence of reliability and construct validity and permit comparison with prior work (Saunders, Lewis and Thornhill, 2019). The questionnaire is organised into four blocks.

The first block captures the independent variables — the features of remote work — using established scales for job autonomy and for the demands and resources associated with remote work, drawn from the job demands–resources tradition. The second block measures the dependent variable, employee wellbeing, conceptualised as a multidimensional construct spanning psychological, social and work-related components; it employs recognised measures of context-free and work-related wellbeing so that the construct is not reduced to a single indicator. The third block measures theoretically relevant mediators and moderators, such as boundary control and perceived social support. The fourth block records demographic and contextual controls, including remote-work intensity, tenure, role and household composition. Attitudinal items use balanced multi-point Likert-type response formats, with a mix of positively and negatively worded items to attenuate acquiescence bias, and attention-check items are embedded to identify inattentive responding.

Full details of each source scale, its number of items and its published reliability are to be tabulated; the design principle is that construct measurement is inherited from validated instruments and adapted only where the remote-work context demands rewording, with any such adaptation itself subject to the pilot described in Section 3.6.

3.5.2 Qualitative Instrument: The Semi-Structured Interview Guide

Qualitative data are generated through semi-structured interviews, a form well suited to exploring how participants understand and account for their experience while retaining enough structure to address the study’s explanatory foci (King, Horrocks and Brooks, 2019). The semi-structured format offers a disciplined flexibility: a common guide ensures that the same substantive territory is covered across participants, supporting cross-case comparison, while open questions and probes allow each participant to lead the conversation into the meanings that matter to them, admitting the unexpected (Kvale and Brinkmann, 2015).

The interview guide is organised thematically rather than as a fixed script. It opens with a broad, non-threatening descriptive question inviting participants to narrate a typical remote working day, which establishes rapport and grounds subsequent discussion in concrete experience. It then moves through thematic areas aligned with the research questions — autonomy and control over work; the negotiation of boundaries between work and home; connection, isolation and support; and the participant’s own account of how remote work has affected their wellbeing over time. Each area is supported by a small number of open lead questions and a repertoire of probes (“can you give me an example?”, “what was that like for you?”) to deepen responses. Crucially, because the design is explanatory sequential, the generic guide is tailored after the quantitative analysis: specific quantitative findings inform bespoke lines of questioning, and where a participant’s survey profile is unusual, the guide is adapted to explore that particular pattern. Interviews are conducted via video conference to match the remote context of participants, audio-recorded with consent, and transcribed verbatim for analysis.

3.6 Pilot Study

A pilot study is conducted prior to main data collection to test and refine both instruments, in line with the principle that piloting reduces measurement error and improves the practical conduct of fieldwork (Bryman and Bell, 2015). Piloting is not a formality but a substantive quality-assurance stage, and its outcomes are documented and reported.

For the quantitative questionnaire, the pilot pursues three aims. First, it assesses face and content validity: a small panel of subject experts reviews the item pool for relevance, clarity and coverage of the constructs, and the instrument is then completed by a modest sample of respondents drawn from the target population but excluded from the main study. Second, it examines item performance and the internal consistency of each multi-item scale, so that problematic or redundant items can be identified before the main survey. Third, it checks practical matters — completion time, routing logic, the functioning of attention checks and the clarity of instructions — since a burdensome or confusing questionnaire depresses response quality. Any wording adapted for the remote-work context is specifically scrutinised here.

For the qualitative strand, one to three pilot interviews are conducted to rehearse the interview guide, to test the ordering and comprehensibility of questions, to gauge the productivity of the probes and to allow the researcher to reflect on and improve their own interviewing technique, including the management of silence and the avoidance of leading questions (Kvale and Brinkmann, 2015). The pilot also tests the recording and transcription workflow. Insights from the pilot feed forward into revised instruments; where changes are substantial, the revised questionnaire items are re-checked so that the instrument entering the field is demonstrably improved rather than merely tested.

3.7 Data Analysis

Consistent with the sequential explanatory design, the two strands are analysed in sequence and then integrated, so that the analytical procedures mirror the logic of the design as a whole.

3.7.1 Quantitative Analysis

Survey data are analysed using statistical software. Analysis proceeds in stages. The data are first screened and cleaned: responses failing attention checks or falling below a minimum completion threshold are removed, missing-data patterns are examined and handled by a principled method appropriate to their mechanism, and the distributional assumptions underpinning the planned techniques — normality, linearity, homoscedasticity and the absence of problematic multicollinearity — are assessed (Tabachnick and Fidell, 2019; Field, 2018). Multi-item scales are then evaluated for internal consistency, with Cronbach’s alpha reported for each and a conventional threshold applied, and, where appropriate, the measurement structure is examined through factor analysis to confirm that items load on their intended constructs.

Descriptive statistics summarise the sample and the central tendency and dispersion of each variable. Inferential analysis then addresses the first research question. Bivariate relationships are examined through correlation, and the core hypotheses are tested using multiple regression, which estimates the unique contribution of each remote-work feature to wellbeing while controlling for the demographic and contextual covariates. Hypothesised mediation and moderation — for example, whether boundary control mediates the effect of temporal flexibility on wellbeing, or whether social support moderates the effect of isolation — are tested using established regression-based procedures for conditional and indirect effects. Effect sizes and confidence intervals are reported alongside significance tests, in recognition that statistical significance alone is an incomplete basis for inference (Field, 2018). The output of this phase is not only a set of tested hypotheses but, critically for the design, a map of which results are strong, weak, surprising or ambiguous, and thus most in need of qualitative explanation.

3.7.2 Qualitative Analysis

Interview transcripts are analysed using reflexive thematic analysis following the six-phase approach of Braun and Clarke (2006, 2022). This method is selected because it is theoretically flexible — compatible with the study’s pragmatist, abductive orientation — and because it is well suited to identifying patterns of meaning across a data set while preserving the interpretive depth the second research question requires. The analysis is predominantly latent and constructionist in emphasis, seeking not merely to catalogue what participants say but to interpret the assumptions and meanings underlying their accounts.

The six phases proceed as follows: familiarisation with the data through repeated reading of transcripts and note-making; systematic generation of initial codes across the full data set; the construction of candidate themes by collating codes into patterns of shared meaning; the reviewing of themes against both the coded extracts and the entire data set; the defining and naming of themes to establish their scope and essence; and the production of the analytical narrative that weaves themes together with vivid data extracts into a coherent argument. Consistent with the reflexive tradition, coding is understood as an interpretive act shaped by the researcher’s engagement rather than as a mechanical procedure aimed at inter-coder reliability; Braun and Clarke (2022) explicitly caution against treating coding reliability as a marker of quality in this variant of the method. The analysis is supported by qualitative data-analysis software to organise codes and extracts and to maintain an auditable record of analytical decisions.

3.7.3 Integration

Integration is the defining act of the mixed-methods design and is pursued deliberately rather than left to emerge (Bryman, 2006; Fetters, Curry and Creswell, 2013). Three integrative techniques are used. Connecting occurs at the sampling stage, where quantitative results drive the selection of interview participants. Building occurs at the instrument stage, where quantitative findings shape the tailored interview questions. Merging occurs at the interpretation stage, where the two sets of results are brought together in joint displays — matrices that array quantitative findings alongside the qualitative themes that explain them — enabling systematic assessment of where the strands converge, complement, expand or contradict one another (Fetters, Curry and Creswell, 2013). The final step is the drawing of a meta-inference: an integrated conclusion that neither strand could reach alone, in which the qualitative accounts furnish the mechanisms and boundary conditions that give the measured relationships their explanatory content. Instances of divergence between the strands are treated not as errors to be reconciled away but as analytically productive puzzles that sharpen the final interpretation.

3.8 Validity, Reliability and Trustworthiness

Because the study combines two methodological traditions, its quality is appraised against the criteria appropriate to each, together with criteria specific to the integration itself.

3.8.1 Quantitative Quality: Validity and Reliability

For the quantitative strand, quality is assessed through the classical criteria of validity and reliability (Bryman and Bell, 2015; Field, 2018). Construct validity — the extent to which the instrument measures the constructs it claims to — is supported by the use of previously validated scales and by the factor-analytic examination of measurement structure. Content validity is addressed through the expert-panel review in the pilot. Criterion and convergent validity are examined through the pattern of correlations among related measures. Internal validity — confidence that observed associations reflect the hypothesised relationships rather than confounds — is strengthened by the inclusion of theoretically motivated control variables, though the cross-sectional survey cannot establish causal direction, a limitation acknowledged in Section 3.10. External validity is served by the stratified sampling and the a priori power analysis, which together support cautious generalisation to the accessible population. Reliability, the consistency of measurement, is evidenced by the internal-consistency coefficients reported for each scale and by the standardisation of administration afforded by the self-completion online format.

3.8.2 Qualitative Quality: Trustworthiness

Applying reliability and validity in their positivist sense to qualitative data would be a category error; the qualitative strand is instead evaluated against the four criteria of trustworthiness established by Lincoln and Guba (1985): credibility, transferability, dependability and confirmability.

Credibility — the qualitative analogue of internal validity, concerning the believability of the account — is pursued through prolonged engagement with the data, the use of verbatim extracts to ground interpretations in participants’ own words, and, where practicable, member reflections in which the plausibility of emerging interpretations is discussed with participants. The maximum-variation sampling and the search for divergent as well as confirming cases guard against a one-sided reading. Transferability — the analogue of external validity — is supported not by statistical generalisation but by the provision of “thick description” (Lincoln and Guba, 1985) of the context, participants and settings, so that readers can judge the applicability of the findings to their own contexts. Dependability — the analogue of reliability — is addressed by maintaining a transparent, auditable record of methodological decisions across the fieldwork, such that the logic of the inquiry could be followed and, in principle, examined by an external auditor. Confirmability — the analogue of objectivity, concerning the extent to which findings are shaped by participants rather than by researcher bias — is served by the same audit trail together with a reflexive journal in which the researcher documents their assumptions and their evolving relationship to the data. Reflexivity is treated here not as a threat to be eliminated but, following Braun and Clarke (2022), as an analytic resource that is examined and made visible.

3.8.3 Legitimation of the Integration

Beyond the two strands, the quality of their combination is itself appraised. Following the notion of inference quality and legitimation in mixed methods (Tashakkori and Teddlie, 1998; Onwuegbuzie and Johnson, 2006), the study attends to whether the sequential ordering of the phases has introduced bias, whether the sample for the qualitative phase adequately represents the quantitative sample, and whether the meta-inference is genuinely warranted by both bodies of evidence rather than by one privileged over the other. The joint-display method makes these judgements explicit and open to scrutiny.

3.9 Ethical Considerations

The research is conducted in accordance with the principles of research ethics and is subject to formal review and approval by the institutional research ethics committee before any data are collected. The ethical framework is organised around the established principles of informed consent, confidentiality and anonymity, protection from harm, and integrity in data handling (Bryman and Bell, 2015; Saunders, Lewis and Thornhill, 2019).

Informed consent is secured at both phases. Survey respondents are presented with a participant information sheet describing the purpose of the study, what participation involves, the voluntary nature of participation and their right to withdraw, and consent is recorded before the questionnaire begins. Interview participants receive a fuller information sheet and provide separate, explicit written consent to be interviewed and audio-recorded, with consent treated as ongoing and re-confirmed at the interview. Confidentiality and anonymity are protected by separating identifying contact details from response data, by pseudonymising transcripts, and by removing or masking identifying particulars in any reported extract; because interview participants are drawn from survey respondents, particular care is taken that the linkage cannot be used to re-identify individuals. Protection from harm is a live consideration because wellbeing is a sensitive topic and interviews may touch on distress; the interview guide is designed to avoid unnecessary intrusion, participants are reminded that they may decline any question or pause at any time, and signposting to appropriate support resources is provided. Data management complies with applicable data-protection law and institutional policy: data are stored securely on encrypted, access-controlled systems, retained only as long as necessary and disposed of appropriately. Finally, the study attends to the power relations and organisational context of workplace research, ensuring that participation is genuinely voluntary and that no employee is disadvantaged by declining, and that findings reported to participating organisations are aggregated so that no individual can be identified by an employer.

3.10 Methodological Limitations

Sound methodology requires the candid acknowledgement of limitations, so that the reader can appraise the conclusions in full knowledge of the design’s boundaries.

First, the quantitative strand is cross-sectional, capturing relationships at a single point in time. It can establish association but cannot, on its own, establish the causal direction or temporal ordering of the relationships between remote-work features and wellbeing; reverse causation and reciprocal effects cannot be excluded. This is mitigated but not removed by the theory-driven specification of the models and by the explanatory qualitative phase, and a longitudinal extension is identified in the concluding chapter as a priority for future work.

Second, both strands rely on self-report, which introduces the potential for common-method variance and for social-desirability and recall biases. The questionnaire design mitigates this through validated scales, balanced item wording and attention checks, and the mixed design itself reduces reliance on any single method, but the limitation remains inherent to a study of subjective states.

Third, generalisability is bounded by the definition of the accessible population. Focusing on knowledge workers in specific sectors secures analytical control at the cost of transferability to manual, frontline or public-sector remote contexts, to which the findings should be extended only with caution.

Fourth, the qualitative strand carries the interpretive subjectivity intrinsic to reflexive thematic analysis. This is managed as a resource through reflexivity and an audit trail rather than eliminated, but a different analyst might have constructed the themes differently; the account is offered as a credible and well-evidenced interpretation, not as the only possible one.

Fifth, the sequential explanatory design is resource-intensive and time-dependent, and the dependence of the second phase on the first means that any weakness in the quantitative results propagates into the qualitative sampling. The pilot study and the a priori power analysis are intended to protect the integrity of the first phase precisely because so much rests upon it.

3.11 Chapter Summary

This chapter has set out and justified the methodological architecture of the thesis. It located the research within a pragmatist philosophy, arguing that the study’s heterogeneous research questions — one explanatory and variance-oriented, one interpretive and process-oriented, one integrative — cannot be answered from within either positivism or interpretivism alone, and that pragmatism’s problem-centred orientation both licenses and disciplines the combination of methods. From this foundation the chapter developed a mixed-methods approach realised through a sequential explanatory design, in which a quantitative survey phase establishes the pattern of relationships between remote-work features and employee wellbeing, and a subsequent qualitative interview phase, whose participants are sampled on the basis of their survey responses, explains those relationships. It specified the population and the stratified quantitative and purposive qualitative sampling strategies, described the validated survey instrument and the semi-structured interview guide, and reported the piloting of both. It detailed the analytical procedures — statistical modelling for the quantitative strand, reflexive thematic analysis for the qualitative strand, and joint displays and a meta-inference for their integration — and appraised the quality of the design against validity and reliability criteria for the quantitative strand, Lincoln and Guba’s (1985) trustworthiness criteria for the qualitative strand, and legitimation criteria for their combination. Finally, it set out the ethical framework and acknowledged the design’s limitations. The chapter that follows reports the findings of the quantitative phase, the results of which shape the qualitative phase presented thereafter.

References

Braun, V. and Clarke, V. (2006) ‘Using thematic analysis in psychology’, Qualitative Research in Psychology, 3(2), pp. 77–101.

Braun, V. and Clarke, V. (2022) Thematic Analysis: A Practical Guide. London: SAGE.

Bryman, A. (2006) ‘Integrating quantitative and qualitative research: how is it done?’, Qualitative Research, 6(1), pp. 97–113.

Bryman, A. and Bell, E. (2015) Business Research Methods. 4th edn. Oxford: Oxford University Press.

Creswell, J.W. and Plano Clark, V.L. (2018) Designing and Conducting Mixed Methods Research. 3rd edn. Thousand Oaks, CA: SAGE.

Crotty, M. (1998) The Foundations of Social Research: Meaning and Perspective in the Research Process. London: SAGE.

Feilzer, M.Y. (2010) ‘Doing mixed methods research pragmatically: implications for the rediscovery of pragmatism as a research paradigm’, Journal of Mixed Methods Research, 4(1), pp. 6–16.

Fetters, M.D., Curry, L.A. and Creswell, J.W. (2013) ‘Achieving integration in mixed methods designs — principles and practices’, Health Services Research, 48(6, Part 2), pp. 2134–2156.

Field, A. (2018) Discovering Statistics Using IBM SPSS Statistics. 5th edn. London: SAGE.

Greene, J.C., Caracelli, V.J. and Graham, W.F. (1989) ‘Toward a conceptual framework for mixed-method evaluation designs’, Educational Evaluation and Policy Analysis, 11(3), pp. 255–274.

Guest, G., Bunce, A. and Johnson, L. (2006) ‘How many interviews are enough? An experiment with data saturation and variability’, Field Methods, 18(1), pp. 59–82.

Johnson, R.B. and Onwuegbuzie, A.J. (2004) ‘Mixed methods research: a research paradigm whose time has come’, Educational Researcher, 33(7), pp. 14–26.

King, N., Horrocks, C. and Brooks, J. (2019) Interviews in Qualitative Research. 2nd edn. London: SAGE.

Kvale, S. and Brinkmann, S. (2015) InterViews: Learning the Craft of Qualitative Research Interviewing. 3rd edn. Thousand Oaks, CA: SAGE.

Lincoln, Y.S. and Guba, E.G. (1985) Naturalistic Inquiry. Beverly Hills, CA: SAGE.

Malterud, K., Siersma, V.D. and Guassora, A.D. (2016) ‘Sample size in qualitative interview studies: guided by information power’, Qualitative Health Research, 26(13), pp. 1753–1760.

Morgan, D.L. (2007) ‘Paradigms lost and pragmatism regained: methodological implications of combining qualitative and quantitative methods’, Journal of Mixed Methods Research, 1(1), pp. 48–76.

Onwuegbuzie, A.J. and Johnson, R.B. (2006) ‘The validity issue in mixed research’, Research in the Schools, 13(1), pp. 48–63.

Palinkas, L.A., Horwitz, S.M., Green, C.A., Wisdom, J.P., Duan, N. and Hoagwood, K. (2015) ‘Purposeful sampling for qualitative data collection and analysis in mixed method implementation research’, Administration and Policy in Mental Health and Mental Health Services Research, 42(5), pp. 533–544.

Saunders, M., Lewis, P. and Thornhill, A. (2019) Research Methods for Business Students. 8th edn. Harlow: Pearson Education.

Tabachnick, B.G. and Fidell, L.S. (2019) Using Multivariate Statistics. 7th edn. Boston, MA: Pearson.

Tashakkori, A. and Teddlie, C. (1998) Mixed Methodology: Combining Qualitative and Quantitative Approaches. Thousand Oaks, CA: SAGE.