• No se han encontrado resultados

Teorías relacionadas al tema

In document FACULTAD DE INGENIERIA (página 35-49)

I. INTRODUCCIÓN

1.3. Teorías relacionadas al tema

The fifth step of Eisenhardt’s (1989a) method is the analysis of data.

Eisenhardt (1989a) breaks the analysis process into two parts; within-case analysis and cross-case pattern search. The within-case analysis gains familiarity with the data and allows preliminary theory generation. The cross-case pattern search forces the researcher to look beyond initial impressions (Eisenhardt, 1989a, p. 533). Eisenhardt (1989a) sees analysing data as the

“heart of building theory from case studies” but warns that it is also the “most difficult and least codified part of the process” (p. 539).

The computer software package QSR NVivo version 2 (www.qsrinternational.com) is utilised to assist in the analysis and data management of this research. The use of computer software in the analysis, particularly the coding, classification and pattern forming from large volumes of data is seen to have clear advantages over manual systems in terms of both speed and rigour. QSR NVivo also acts as a case study database, to underpin the reliability of the research.

6.2.5.1 Grounded

While this research begins as close as possible to ‘no theory under consideration’ specific grounded theory techniques (Glaser & Strauss, 1967;

Strauss & Corbin, 1998) are not utilised in this analysis. In Eisenhardt’s later work in conjunction with Graebner (2007), she specifically distances herself from the grounded theory method. Eisenhardt and Graebner (2007) see confusion around the multiple meanings of ‘grounded theory building’ and

106

view ‘grounding’ as the strength of the theory’s empirical grounding rather than the “specifics of the theory-building process” and suggest avoiding the use of the term “grounded theory building” (p. 30). As such, this analysis uses Eisenhardt and Graebner’s (2007) meaning of ‘grounded theory building’ as simply “creating theory by observing patterns within systematically collected empirical data … [including] some notion of recursively iterating between (and thus constantly comparing) theory and data during analysis, and theoretically sampling cases” (p. 30). This pattern of observation forms the basis for analysis (Chapters 7 and 8).

6.2.5.2 Within-case Analysis

Eisenhardt (1989a) does not suggest a standard format for within-case analysis, but encourages the researcher to allow “the unique patterns of each case to emerge” (p. 540). Within-case analysis focuses on developing constructs and relationships within the co-operative case, through case write-ups, coding and pattern coding to inductively allow constructs to emerge from the data (Graebner & Eisenhardt, 2004).

Within-case analysis, initially involving descriptive detailed case write-ups, is utilised to help the researcher deal with the volume of data and to “become intimately familiar” with each case (Eisenhardt, 1989a, p. 540). After “getting a feel” for the transcribed interviews15 initial case study write-ups are completed. These write-ups utilise many sources of data, including interviews and archival data, particularly annual reports, company constitutions and published company documentation. Case familiarity assists in later cross-case comparisons and encourages the resultant theory to be firmly grounded in the empirical data (Eisenhardt & Graebner, 2007).

Deeper within-case analysis commences with the initial interview transcriptions coded sentence by sentence. Initial codes are sometimes

15Analysis of interviews began after the return of transcribed interviews from participants and any corrections, deletions or additions are incorporated.

107

inductively derived from the respondent’s frame of reference and on occasion field notes, including contact summary sheets, are used to highlight initial codes. This again assists in grounding the resultant theory in the empirical data. Codes are assigned “clear operational definitions” (Miles & Huberman, 1994). Codes, once created, are used in the coding of subsequent interviews.

During analysis some codes prove superfluous and are dropped; others flourish in light of their fit and descriptive power as data collection and analysis progresses in tandem. As theory development is “a process of describing phenomena at increasingly higher levels of abstraction” (Zikmund, 2003, p.

41) codes are then grouped together into common categories (or themes). To develop categories, memoing is used to question data and clarify ideas about the relationships between codes (Miles & Huberman, 1994). These categories are labelled and defined. The empirical data coded under a category is compared and contrasted with data previously coded under the category. In the iterative process entire interviews are regularly recoded as different thinking emerges. During the analysis, categories grew or dissipated and merged as evidence supported or challenged the categories to capture higher levels of abstraction. In this manner, initial descriptive codes became increasingly interpretive.

In order to move to a more inferential and explanatory level the initial coding, memoing, categories and writing are employed to induce patterns of behaviours, or pattern codes and constructs (Miles & Huberman, 1984).

Pattern codes revolve around “themes, causes/explanations, relationships among people, emerging constructs” (Miles & Huberman, 1994, p. 70).

Initially, the roles or what boards actually do is utilised in pattern recognition, however the importance of processes, relationship dynamics and decision making process became increasingly apparent in the understanding of co-operative board roles. Once pattern codes are developed they are employed on the next lot of empirical data to test for fit and explanatory power.

In line with the method, data collection and analysis are carried out simultaneously in an iterative way. New data is continually tested against the

108

emerging interpretations and explanations as they develop to enlighten and extend the emergent theory. “Serendipitous findings” and provisional analysis of the data inform further data collection, in particular, revisions to the interview schedule to fill in gaps and pursue new leads. As an example, the initial focus of the research is on board roles, but the analysis and subsequent data collection suggest that an understanding of relationships both within the board and between key stakeholders - the board, management and supplier-shareholders - are an important aspect of understanding it.

As data is continuously reduced, care is taken to return to the original transcripts and documents to ensure that theorising remains grounded in the empirical evidence. Initially, all explanations and conclusions emerging are held lightly “maintaining openness and scepticism” (Miles & Huberman, 1994, p. 11) until rival explanations are discounted and to prevent premature theoretical closure. Through the inductive process the categories are strengthened and reduced by seeking corroborating and disconfirming evidence until the weight of evidence supporting concepts16 emerge as an abstract of reality. Concepts are the primary building blocks of theory (Llewelyn, 2003; Zikmund, 2003). Concepts “subsume a mountain of particulars” and by determining the relationships between them lead to a conceptual framework (Miles & Huberman, 1994, p. 18).

6.2.5.3 Conceptual Framework

As patterns start to appear networks are formed to try to make the analytic progression to understand the explanations behind the particular patterns. From the “what” and “how” to the “why” (Miles & Huberman, 1994, p. 90).

Qualitative interview data is important in gaining an understanding of these emergent relationships. Diagrams, including boxes and arrows, are then used to try to establish relationships between variables of interest. The patterns are then compared against the transcripts to ensure they are not drifting from the

16 Zikmund (2003) defines a “concept (or construct) is a generalized idea about a class of objects, attributes, occurrences, or processes that has been given a name” (p. 41). The term concept is used in these findings. Eisenhardt (1989a) seems to use the terms concepts and constructs interchangeably.

109

empirical evidence. This leads to a conceptual framework (displayed in Figure 7.1, Chapter 7) which is at the heart of building theory, showing “how the variables are connected, how they influence each other” (Miles & Huberman, 1994, p. 91). Due to its inductive nature, the model emerges on a piecemeal basis with the initial version “amended and refined” as it is “tested against empirical events and characteristics” (Miles & Huberman, 1994, p. 155).

6.2.5.4 Cross-case Patterns

The second step in analysing data in Eisenhardt’s (1989a) framework is searching for cross-case patterns. Tactics suggested for cross-case analysis can involve looking for and listing similarities or differences across categories, dimensions, between matched pairs or by data sources (Eisenhardt, 1989a, p.

540). Cross-case analysis began at the conclusion of within-case analysis to begin to draw conclusions on the entire study (Eisenhardt, 1989a; Yin, 1994).

Similarities and differences among cases are noted during data collection and within-case analysis, although further analysis is deferred until all cases have been analysed “in order to maintain the independence of the replication logic”

(Graebner & Eisenhardt, 2004, p. 373).

Pattern matching across-cases is utilised. Parkhe (1993) sees pattern-matching as comparing an empirically based pattern induced from one case with other cases, so that they either match or are predictably different than the initial case patterns (pp. 249-250). In the analysis of qualitative data “gross matches or mismatches” are sought (Yin, 1994). Pettigrew (1997) also suggests cross-case pattern comparison to “find the underlying mechanisms which shape any patterning in the observed processes” and notes teasing out these mechanisms

“represents one of the greatest inductive challenges for process scholars and an area of intellectual challenge which is as difficult to describe as it is to achieve and publically justify” (p. 339). The data are looked at in many divergent ways to force the researcher beyond his preliminary impressions (Eisenhardt, 1989a). Rival explanations for patterns are tested within and across-cases to allow new insights and to “improve the likelihood of accurate and reliable theory” with a close fit to the empirical data (Eisenhardt, 1989a, p. 541). Rival

110

explanations are either countered, thus strengthening the developing model, or cause revision of the empirical model and thus firmer grounding of the findings. Eisenhardt’s (1989a) suggestions of cross-case searching for patterns of “within-group similarities and intergroup differences” from dimensions suggested by the “research question or by existing literature” are utilised to strengthen internal validity (p. 540).

In summary, the empirical data is analysed both within-case and across-case to inductively develop a theory of the roles of New Zealand and Australian co-operative boards grounded in the empirical data.

In document FACULTAD DE INGENIERIA (página 35-49)

Documento similar