Soon after each interview, I conducted the ‘time-consuming and soul destroying task’
(Gorman & Clayton, 2005, p. 137) of transcribing the tape-recorded interview data.
The total amount of content from the interviews came to approximately 18 hours of recorded data and 326 pages of transcribed text.
In the first three sub-studies (studies I–III), I used the interview data, approaching the data from a different perspective for each, in line with the respective research questions and research interests. The process of analysis followed the same basic principles in all three, though. I analysed the data by applying qualitative methods.
Generation of codes representing the categories of data can be drawn from previous literature and from interview themes. This kind of concept-driven coding differs from data-driven coding, wherein the codes are derived from the data without many preconceptions. Also, a study can apply both approaches. (Gibbs, 2007, pp.
44–46.) In the study reported upon here, previous pieces of literature were utilised as sources of inspiration for coding of the data.
Characterisation of the data (Savin-Baden & Major, 2013) was a one-time exercise. The interviews were recorded and transcribed verbatim, though verbal tics and murmuring were excluded from the transcripts. I immersed myself in the data by reading the transcribed text and listening to the recordings several times. For each of the sub-studies (studies I–III), the coding of data to categorise the text, identification of items from the data, and conversion of the codes and categories into themes had distinctive nuances, presented later in this chapter.
In the process of handling and analysis of the data, the data-analysis software ATLAS.ti was consistently used. ATLAS.ti is a tool that supports the organisation and analysis of qualitative data. Computer programs offer advantages for qualitative data analysis, such as easing the burden of the process of writing and rewriting;
supporting sorting, referencing, and coding; and facilitating creation of statistical tables and graphics (Gorman & Clayton, 2005, p. 220).
In Study I, firstly, the appropriate parts of the dataset were coded and categorised under the themes in line with the research questions stated. Next, focused sub-codes were identified and added to the categories. Then, a hierarchical list of codes was produced in parallel with the reading process. The coding system was created to aid in understanding the data and conceptualising the codes (Friese, 2012, pp. 122–123).
Finally, the data behind the codes were examined one by one and in parallel with each other in light of the research questions.
In Study II, procedures presented by Savin-Baden and Major (2013, pp. 420–433) were used to some extent. After characterisation of the data was immersion in the data. Next, the dataset was divided into segments, and the resulting chunks were denoted in ATLAS.ti. The data then were coded for categorisation to present a framework for analysis. Difficulties and the methods for handling them were identified from the data. After that, these were linked together and elucidated via graphical illustrations produced by ATLAS.ti (Friese, 2012, p. 216).
In Study III, elements from the transcripts were categorised by means of codes.
The analysis strategy was again supported by ATLAS.ti. Code-and-retrieve strategy (Coffey & Atkinson, 1996, p. 170) was implemented in the data’s organisation. In generating and using ideas and in generalising, the analysis was informed by writing analytic memos and creating lists. Recordkeeping professionals’ perceptions were exemplified by translations of excerpts from the interviews.
In Study IV, facet analysis was applied. With this analysis method, the titles used in functional classification systems were organised into facets. In facet analysis,
analysis since normally that level is used for classifying records in Finnish registration practice. For selecting a representative sample from each of the classifications’ class names, probability sampling techniques were applied (Pickard, 2007, p. 61). The full sample size was 315 class names, 105 from each classification. An Excel spreadsheet was used in listing the titles and in the analysis.
Five suitable facets were created. They were derived a priori from previous research (e.g., Alberts et al., 2010; Sabourin, 2001; Shepherd & Yeo, 2003) that described functions and how one can create functional classifications. The following facets were used:
1. Actor (who / which part of the organisation is acting) 2. Action (what is happening / what is being done)
3. Object of Action (the object / target / receiving end of the action) 4. Subject (what subject matter / theme is dealt with in the class)
5. Object of Documentation (the outcome/record articulated in the class name) The numbers of entities under each facet were summed in Microsoft Excel. The class names varied in their nature and structure. Hence, there was variation in their distribution across the facets. Some of the class names fitted one facet, while others included elements from more than one facet. For presentation of the findings, tables and simple bar charts were created. Also, the results were qualitatively described.
4 Findings
The original publications, covering studies I–IV, constitute the core output of the thesis project. In this chapter, a summation of the key findings from studies I–IV is presented. The individual sub-studies are addressed in chronological order by the time when they were conducted and written up, not that of publication in journals.
In Study I, recordkeeping professionals’ understanding of functional classification, their perceptions of the purposes it serves, and their justification for the approach were addressed. Study II focused on the difficulties that recordkeeping professionals have faced with functional classification systems and how they were handling those difficulties. In Study III, recordkeeping professionals’ perceptions of other users of functional classification systems in Finnish public-sector organisations were explored. Finally, Study IV examined the class names used at the lowest function level of the hierarchy in these organisations’ functional classification systems. After the findings of the respective sub-studies are discussed, below, an integrative summary of the findings from all four is presented.