3.6 Approaches to Data Analysis
3.6.1 Constructivist grounded theory
A constructivist grounded theory approach to research is an inductive method for theory generation, drawing theory from the data rather than the data serving to verify a preconceived hypothesis or theory, working from the theory
Punch, 1998; Saldaña, 2009). “Theory emerges from … and is said to be ‘grounded’ in” the data” (Whipp, 1998, p. 52). The data collection and
analysis occur simultaneously, as a means of refining and shaping the emergent theories and ensuring these theories are grounded in the data (Charmaz & Mitchell, 2002; Corbin & Strauss, 1990). Within this study, a simplified constructivist grounded theory coding analysis is employed, in which four main coding stages are used, to construct the codes, categories, themes and discourses. These four main stages are termed initial, focused, axial and theoretical coding, respectively (Charmaz, 2006; Saldaña).
During the initial coding stage, the data are examined in close detail (line-by-line), to construct the first stage of codes. These codes are kept active, to emphasise the focus on what is happening in the data (Charmaz, 2005, 2006). Using active codes “curbs our tendencies to make conceptual leaps and to adopt extant theories before we have done the necessary analytic work” (Charmaz, 2006, p. 49; emphasis in original). These active codes are the “beginning” stage of the coding analysis that enables the researcher to “fracture or split the data into individually coded segments” (Saldaña, 2009, p. 42). These individually coded segments represent the first stage of the coding, from which patterns and meanings in the data can eventually be identified.
The first three codes constructed from each of the two codes data types in this research are:
Interview transcript codes 1-3:
Identifies differences in teachers’ methods of test administration (code 1)
Presents tests as isolated activities, no preparation (code 2)
Identifies tests used as diagnostic tools to identify student weaknesses
(code 3)
Testing session transcript codes 1-3:
Presents tests in a positive way (code 1)
Presents tests in a negative way (code 2)
Presents tests as useful/helpful for students and learning (code 3)
Figure 3-1. Codes 1-3, interview transcripts and testing session transcripts.
Codes serve a deeper purpose than merely being tools or terms used by the researcher for differentiating distinct or related meanings, or being helpful labels given to units of meaning identified as important in the data (Dey, 1999). These initial, active codes begin the process of categorising and
conceptualising the patterns of action and meaning. These codes are read and re-read through constant comparison against the data and the other codes, to ensure they conceptualise what is happening in the data. This in turn assists in grounding future coding stages in the data. Initial coding is “open-ended” that allows the “researcher to reflect deeply on the contents and nuances of your data and to begin taking ownership of them” (Saldaña, 2009, p. 81).
The initial codes are the result of the researcher’s examination of the participant interview and testing session transcript data. These data were examined for what Charmaz (2006) refers to as “sensitizing concepts” (p. 17). These sensitising concepts in this particular coding analysis reflect the research questions: standardised literacy testing value and validity; participants’
perceptions and interpretations of standardised literacy testing; approaches to test administration; and patterns of classroom interaction as related to testing.
It is not only these concepts that are identified in the coding stages, however, as any emergent points of interest are followed, then later refined and revised.
Once the data are coded, the initial codes are concentrated and grouped together into categories, to show the processes and patterns identified in the initial codes. In this research, the stage of sorting codes into categories is termed focused coding. Focused coding is more analytical and theoretical than the initial coding because it requires the analyst to reflexively interact with the data and the active codes. The categories constructed from the focused coding are “more directed, selective, and conceptual” than those constructed from the initial coding phase, and enable the synthesis and examination of “larger segments of data” (Charmaz, 2006, p. 57). This is more theoretical and conceptual than the initial coding as this stage necessitates critical thinking about what the initial codes mean and how they relate.
The first categories and the first three codes that constitute them from each of the two codes data types are:
Classroom control (interview category 1)
Refers to SLT administration rules (code 14)
Stresses/identifies/refers to the
role/importance of classroom control in lesson and testing sessions (code 15) Refers to poor student behaviour – influential upon teaching and administration (code 17) Pragmatics of standardised
literacy testing administration, and administrational uses (test session category 1)
Over-identifies administration as feature/issue of tests (code 16)
Refers to the administrative uses of test data (code 17)
Reads aloud test instructions (code 36)
In this study, as already stated, frequency counts and calculations of codes in the data were undertaken in this and the following stages, and
comparison of themes and behaviours was also included. It is at this stage that the coding begins to highlight the “processes” in the data (Charmaz, 2006, p. 51).
Processes and categorisation are refined and revised in the third coding stage, which further conceptualises the data. The third step of the
constructivist grounded theory coding analysis employed in this research is here termed axial coding, so called because the coding levels and data are re- examined and conceptualised around the core, central ‘axes’ (Saldaña, 2009). This coding stage deconstructs the seven participant interview categories and nine testing session categories, and effectively “describes a category’s
properties and dimensions and explores how the categories and subcategories relate to each other” (Saldaña, p. 151).
The categories are broken up and examined, to make the data “denser, more complex and more precise” (Charmaz & Mitchell, 2002, p. 515). This stage allows for the identification of the themes from the data and the codes, in order to conceptualise and represent the “conceptual elements of [the] theory” (Glaser & Strauss, 1967, p. 36, cited in Dey, 1999, p. 48). At this stage, the coding from both main data types is synthesised, so the themes represent the axial coding of both data types already patterned into codes and categories. The first of these themes and the categories that constitutes it is:
Maintenance of control, behaviour management, order (theme 1)
Classroom control (interview category 1)
Teacher as authority (test session category 9)
Figure 3-3. Theme 1 with categories.
Following the construction of the themes, the properties and parameters of the themes are illuminated and clarified in Chapter Five through the process of examining how the themes address the research questions. It is from this stage of the analysis that the dominant discourses are drawn. The theoretical coding stage breaks down the themes to clarify and question the broader patterns of meaning-making, identity, power and knowledge regarding
standardised literacy test administration for these participants and at this school site are clarified. At this stage of the coding, the analysis “progresses toward discovering the central/core category that identifies the primary theme of the research” (Saldaña, 2009, p. 151), which in this research means the dominant discourses. This adapted theoretical coding stage is undertaken through the process of examining the themes through the lens of the research questions.
‘Theoretical saturation’ is the term used for the situation in which “the data are sufficient to make and justify an interesting argument” (Phillips & Hardy, 2002, p. 74). In this study, theoretical saturation was reached when the identified codes and themes had ‘saturated’ the identifiable in the data by the researcher; the codes and themes were sufficiently thorough, detailed and identifiable (through quotes and examples) so that further attempts to code and compare the data highlighted no new themes regarding the topic. This process
of iterative coding involves returning to data and codes to ensure the thoroughness, saturation and grounding of coding. This process aided the identification of theoretical patterns within and across the levels of coding, and helped to highlight key themes and dominant discourses, particularly through the later coding stages during which the codes were refined, compared, revised and broken down to reveal more complex patterns within the data.
As an iterative coding process, the data and levels of codes are read and re-read to deepen understanding and appreciation of the content, patterns and themes within the data. This iterative nature of constructivist grounded theory coding qualifies it as a constant comparative approach (Charmaz, 2005, 2006). The comparative and iterative nature of this analysis, which allows
reinterpretation and rediscovery, together with a poststructuralist-influenced questioning of data and of the researcher’s self (reflexivity), helps to “make fundamental processes explicit, render hidden assumptions visible, and give participants new insights” (Charmaz, 2006, p. 55).
The constructivist grounded theory coding stages allow researchers to identify themes and issues throughout collected data, locate points of
comparison and group together theoretically similar codes and categories. In this way, the coding meant that in this research the data were thoroughly examined for themes and issues, participants’ interpretations, arguments and justifications, and test-related behaviours, processes and types of interactions.
The researcher employing this analysis approach uses memos, recorded alongside the data and coding, which can be utilised to determine and analyse the deeper context and meaning of the data, codes, categories and thematic
areas. This research adopted the use of memo records during the data generation, collection and analysis stages. Memos allow the researcher to record interesting theories and ideas about each segment of data or the data as a whole, observations that would otherwise be forgotten or lost. Memos also serve as a reminder of these initial theories and ideas and of the context from which they were taken, so redrawing the researcher into the data and the participants’ realities.