Data analysis

As the study applies a mixed method design, data will be analysed using both quantitative and qualitative techniques. The interview data and quali-tative data collected via the Q-sorts will be analysed to search for key themes and patterns. Particular attention will be given to respondents’

comments on perceived benefits of failure learning and strategies to un-learn unsuccessful behaviour. In addition, the subjective data from the Q-sorts will be quantified by application of the Q factor analysis. Finally, data from the Social Style questionnaire and the failure learning arche-types yielded by the Q methodology will be analysed using cross-tabu-lation and association tests (Bühl & Zöfel, 2002). Procedures of data analysis are illustrated next.

Interpretative Phenomenological Analysis

As a nascent approach to phenomenological research, IPA provides an accessible qualitative research method (Larkin & Thompson, 2012). The authors describe the outcome of a successful IPA study as bridging the elements of “giving voice” (to the participant’s narrative and their reflec-tion on the researcher) and of “making sense” (through the interpretareflec-tion of the participant’s account by using psychological concepts). Further-more, the authors outline that finding the right balance between these key components requires substantial time and effort. That said, the method makes no claim to objectivity, rather it is emphatically inductive and idi-ographic. Therefore, the analysis starts with a thorough, detailed examina-tion of one case, and thereafter moves to the careful analysis of subse-quent cases (Cope, 2011; Smith et al., 2009). The extensive, thorough and rigorous procedure of analysis illustrated in table 10 in the dissertation should be convincing enough to show the ability of hermeneutic phenom-enological studies to make meaningful theoretical contributions to

entre-acker-softwar acker-softwar

acker-softwar acker-softwar

preneurial research (Berglund, 2007; Conklin, 2010; Cope, 2011). The analysis has been based on transcriptions of the audio-recorded data from the interviews. In total, 14 interviews have been carried out and all of these interviews have been transcribed, using the services of a profession-al scientific transcription provider. The recordings in totprofession-al consist of more than 15 hours interview time, with an average length of 64 min (ranging from 48 to 97 min) and resulted in 308 transcribed A4 pages which were taken into account for further analysis based on the process illustrated in table 10 of the full dissertation. The results of the analytical process are shown in the following chapter 4.

Q methodology

The study uses Q methodology to conduct a hybrid qualitative and quanti-tative exploration of failure learning. The general five-staged process of a Q methodology study has been illustrated already in section 3.3.3, this section highlights the procedures carried out in the fifth and final stage of the method.

Data from the 28 Q-sorts were entered in an Excel spreadsheet that was imported into the R platform, a free software environment for statistical computing and graphics. Zabala (2014) developed the package qmethod that surpasses other existing, free-of-cost available Q software in many ways, especially by the step-by-step analysis that helps the researcher to fully understand the process.

For the statistical analysis, the Q set becomes the “subjects” and the indi-vidual Q-sorts (carried out by the participants, presenting their indiindi-vidual viewpoints) become the “variables” (Sinclair, 2019). That allows for a correlation of individual viewpoints that cluster together to similar opin-ions or standpoints. Factor extraction in the qmethod package applies a

acker-softwar acker-softwar

acker-softwar acker-softwar

Principal Component Analysis (PCA) and the extracted factors are vari-max-rotated to produce the maximum differentiation. The selection of factors is done iteratively, using both the researchers’ theoretically in-formed judgement and loadings that maximise both the number of state-ments that have significant loading onto the factors and number of partic-ipants accounting for the factor (Watts & Stenner, 2012). Findings of the statistical analysis are provided in chapter 4.

Failure learning association tests

Each of the 42 participant profiles consists of both self-assessment and third-party assessment of measures of social style and versatility (see sub-section 3.3.3) which are based on the Social Style Profile – Enhanced (SSP-E). All statistical analyses were carried out by Tracom and reports provided by the service organisation have been used for cross-tabulation and descriptive statistical analyses with several association tests by appli-cation of the statistical software IBM SPSS Statistics 25. Results of the analyses are to be found in chapter 4. Furthermore, qualitative data col-lected during the individual participant debriefs are taken into account for the qualitative studies (see sub-section 3.3.3). These findings are selec-tively presented in the respective IPA and Q-Methodology finding section in chapter 4.

Compilation of the mixed method study

The aim of the study is to broaden our understanding about learning in the aftermath of entrepreneurial failure under consideration of behavioural pattern in social interactions. Based on the narratives of entrepreneurs about their lived experience of entrepreneurial failure and their sense-making of the crucial life event, Q methodology, a hybrid research tech-nique has been applied for qualitative and quantitative exploration of

acker-softwar acker-softwar

acker-softwar acker-softwar

learning from failure. Thereafter, statistical analyses were carried out to search for associations between failure learning opinion groups and be-havioural pattern measured by the SSP-E questionnaire, the Social Style assessment instrument. The triangulation of both quantitative and qualita-tive data does not only allow for a deeper understanding of the partici-pants’ learning strategies, but additionally enhances the validity of the research findings (Bryman & Bell, 2007; Stokes, 2011). All findings will be presented in chapter 4 and further discussed in chapter 5. Figure 4 summarizes the compilation of the data analysis.

Figure 4 Compilation of the data analysis

Source: own illustration, based on Schönbohm & Jülich (2016)

ILKAHEINZE PAGE39

Triangulation Qualitative Analysis:

IPA study

Hybrid Analysis:

Q methodology

Failure Learning

Quantitative Analysis:

SSP-E associations

acker-softwar acker-softwar

acker-softwar acker-softwar

4 Research findings

This chapter presents the results identified through the data analysis from the IPA interviews, the supporting document analysis, the Q-sorts and the calculation of the associations between learning strategies and behaviour-al pattern as described in the previous chapter. Structured behaviour-alongside the units of analysis defined in chapter 3, the chapter provides answers to the five research questions (see sections 1.2 and 3.2 respectively). Section 4.1 illustrates an short overview of findings obtained from the interpretative phenomenological analysis, the approach requires working through mul-tiple levels of constructing, de-constructing and clustering emergent themes. Section 4.2 presents the findings on the learning archetypes de-termined by the qualitative and quantitative analysis of the Q sorts. Sec-tion 4.3 introduces the outcomes of the quantitative analysis of learning archetypes in regard to their association with behavioural aspects such as social styles and versatility. The chapter is summarised in section 4.4.