Developing an analytic procedure: using Nvivo 8 to serve the analysis and development of constructive grounded theory

Several challenges were encountered in employing Nvivo 8 to serve the analytic work

of constructive grounded theory. Although each challenge usually came along with another month of trial on a new analytic plan, the journey to searching solutions

helped me to clarify the nature and process of the grounded theory analysis I conducted. In order to illustrate the final analytic procedure of this study as well as to

map a better use of Nvivo 8 in future studies, a list of analytic procedures are demonstrated in this section.

In summary, it was found that both analytic media – paper and computer-assisted

software – had their own strengths and weaknesses in terms of interpretation, familiarity with the original text and data management. Hence, in the end, this study

did not favour either approach; instead a cooperative use of both seemed to be appropriate. In attempting to do so, firstly, the figure below was compiled to compare

Table 3.02 Comparison between coding by different software Coding and analysis by

Aspect Paper (or Microsoft Word) Nvivo 8

Length of open-codes Open-codes are usually short (due to limited space on paper).

Open-codes can be made either short or as long as needed.

(In the case of this study, open-codes should be long enough to carry the researcher’s interpretation and key information of the selected texts.) Familiarity with the link

between produced codes and the original texts

Texts are normally viewed whenever codes are viewed.

Texts not viewed unless

double-clicking on each code.

Renaming codes Name can be edited on paper. Name can be edited by computer.

Making changes to the coverage of the texts

coded/imputed

There is normally no fixed boundary for coded texts, unless texts are highlighted or moved to categories.

It is easy to add more texts on any produced codes. Inputted texts can be removed from the code.

Management on codes and its text when moving

to build tree categories

Extra work (copy and paste) needed on recording links or storing related texts in order to carry the texts with the code.

No extra work needed. Selected texts already attach to the produced code, and will follow the move of the code.

Memo writing Created at the end of texts or in another file.

Created as a Memo that could link to one code or file. Texts within memos could also be coded.

Building relationships (category tree) between codes, sub-categories and

categories

Each code or category needs to be moved one by one. Might require extra work to drawing categories tree on another file.

Sets of codes and categories could be moved simultaneously. Relationships between codes can be presented when it is edited.

Transparency about the theory developing process

Not easy to trace or demonstrate where the developed theoretical framework is coming from which sub-categories, codes or text, because data are too multiplicitous to be managed within one file.

Categories, sub-category and codes clearly displayed in layers (tree codes). Texts are attached with each code, or categories. Related memos can also be linked to the framework.

According to the features illustrated above, it seemed that analysing on paper was better used in the early stage in terms of becoming familiar with the text and quickly

identifying key segments; whereas Nvivo 8 was more effective in the later analytic stages such as data management and developing theoretical frameworks. In search of a

better use of both, the data analysis procedure of this study was formed and involved: 1. Carefully inspecting each word and line of data on paper (Microsoft Word) to

highlight important segments and conduct open coding with short open-codes. 2. Short memos recorded on paper (Microsoft Word) with different colours. This was

for later identification of a key incident or development of a category (Figure 3. 06). Interview data were printed for later use.

3. Conducting initial coding in Nvivo 8 with the assistance of printed data. Researcher’s interpretation of each segment was coded as (one or more than one)

open-code using longer key words, sentences or both.

4. When a segment was complex or contained multiple meanings, firstly, moving

forward and backward from the text in order to find out what was the participant’s concern; and secondly, if the segment contained multiple meanings, it could be

coded into different open-codes (in Nvivo 8).

5. Once an interview transcript was completely coded, the participant’s name was

added in front of each open-code so that each code was more identifiable in later analysis. In this study, produced open-codes were copied to both life history and

cross-case theme sections and analysed respectively (in Tree Node of Nvivo 8). 6. Conducting focused coding right after each interview or participant’s data was

finished. This was to prevent the researcher from dealing with too many unorganised codes at a time and develop premature views of the data in the early

stages of the research process. In this stage, codes, sub-categories and categories were linked to form a category tree (in Tree Node of Nvivo 8).

7. After some key factors were identified through focused coding, more information was added in front of each open-code when conducting the remaining interview

data. This was to facilitate categorizing or sub-categorizing when conducting focused coding later. (Figure 3.07)

8. After all the data were analysed, if a category or theme was formed but lacked some participant’s evidence, the researcher went back to the original interview

data to search if any data (1) had been missed and not coded or (2) could be interpreted from different angles in order to fill the gap. (In this case, it was

suitable to use key-word searching in either Microsoft Word or Nvivo 8).

9. Developing a theoretical framework that illustrated all key factors and its

relationship. This framework was different from the use of the category-tree. Whereas the category-tree contained all data, the framework drew key factors out

from the data and displayed the relationships between them. In order to demonstrate relationships among factors, open-coding was also used to record

connections within the participant’s telling. For example, a participant’s belief about PE teaching was often found to be connected to their in-service training.

This could be found in responses such as: I liked to do it in this way… because I had similar experience back in my PETE training. In this case, this kind of

statement was coded as a link-code to support the relationship. A theoretical framework for this process can be seen in chapter 8, Figure 8.1.

10.Writing up findings was the final stage of data analysis. When writing each section of the findings (theme or issue under the theme), the researcher needed to put the

produced codes in order by further sub-categorizing them in as many layers as needed. The relationship in each theme and category formed the writing structure

Figure 3.06 Making short codes or memos on paper

Figure 3.07 Making codes by using Nvivo 8

Name Category or key factor Content of open-code Other notes Highlight of important texts Open-codes Memos or notes