Definition
The analysis of textual data aided by computer software which has been designed to support the analyst with the storage, coding and systematic retrieval of qualitative data.
Distinctive Features
The social science disciplines have seen a growing interest in the use of computer-assisted qualitative data analysis (CAQDAS) since its introduction in the mid-1980s. A number of software programs have been developed: popular programs include NVivo, NUD*IST (now N6), ATLAS/ti and The Ethnograph.
In essence the principles of computer-assisted data analysis are similar to that of non-computerized analysis. Having obtained textual data by whatever method – interview or focus group transcripts, fieldnotes or documentary sources – the researcher examines the data for emerging themes. Sections of text are then marked or ‘tagged’ with particular codes with usually more than one code being assigned to a given piece of text. The researcher then scrutinizes data held within each code and re-codes the data by creating sub-categories of codes. This iterative process allows the indexing, modification and elaboration of data into a tree-like structure where the branches represent progressively more fine-grained analysis.
Although the process of data coding is similar whether performed by hand or computer assisted, CAQDAS really shows its true value in the context of data searching and retrieval. Qualitative data analysis packages can assist in a num- ber of ways. For example the computer can perform a textual search for a partic- ular string of characters and thus retrieve all data that contains the occurrence of words or phrases. Index searches are another common method of data retrieval and may be simple or can be made more elaborate with the use of Boolean search operators. Therefore a researcher can request the software to retrieve all data coded at code A and code B, but not at code C. By conducting index searches such as this the researcher is able to test the robustness of the coding tree and seek pat- terns or themes in the data. This process also supports the researcher as he or she begins to develop and test hypotheses from the data. Analytic memoranda may also be attached to codes. These may provide a description of the themes of the code or analytical hunches about the data held within it.
As mentioned previously, the basic principles of textual coding and searching are similar whether the researcher uses computer-based or manual
Computer-Assisted Data Analysis
data analysis. However, computer analysis can offer a number of advantages over manual methods (Coffey, Holbrook and Atkinson, 1996; Seale, 2000). Perhaps the most significant benefit is that computer analysis has negated the need for the time-consuming task of cutting, pasting and sorting textual data. What the researcher once did by hand, equipped with piles of paper tran- scripts, coloured pens and a pair of scissors, can now be done by computer, therefore enabling the task to be conducted much more quickly. This is of clear benefit to researchers because the time saved can be invested in other facets of data collection and analysis or can enable researchers to include more data in their projects. A second perceived advantage of computer-assisted analysis is that the data searches will tend to be more rigorous, systematic and comprehensive. While a researcher conducting a manual search may be tempted to truncate the search once they have found enough data to provide evidence for their hypo- thesis, or have found a worthy quote or anecdote to illustrate their point, a computer search will execute the whole task until all occurrences of data have been found. A third advantage of computer-assisted analysis is that computers are able to process much more complicated tasks than the human brain. This has meant that increasingly more advanced Boolean search operators can be used, for example ‘search for data assigned code A but not proximal to code B’. A final advantage of computer-assisted analysis is that it can assist with team analysis. Researchers working in large teams, perhaps even based at different sites, can analyse their data in isolation and then merge and share their data with other team members. Researchers may also code the same data indepen- dently thus corroborating each other’s interpretations of the data.
More recent advances in the computer-assisted analysis programs have heralded the developmental use of hypertext linkages (Coffey, Holbrook and Atkinson, 1996). This technique allows the readers of the research to click a high- lighted icon within the research report and be routed to a hyperlink that may either be a section of the original textual data, a picture or sound file, or per- haps a summary of the respondent’s demographic characteristics. The hyper- links to original data enable the reader to explore the concepts in as much detail as he or she wishes and can provide illustrations of the analyst’s theories thus enabling a more transparent interpretation of the data.
Examples
Buston’s (1997) paper offers a practical account of how CAQDAS may be applied. The paper, based on interview data from chronically ill young people, offers a step-by-step guide to how the data were introduced into and indexed within the NUD*IST programme. The paper also provides an overview of some of the main capabilities of the NUD*IST software including how base data (the
demographic characteristics of respondents) can be included and utilized in the analysis and how indexed data can be retrieved. The paper concludes with a discussion of the main methodological debate around CAQDAS software, that is, whether the software can influence how the analysis is conducted.
An example of the use of hypermedia within computer-assisted data analysis is provided by Dicks and Mason (1998) in their study of a coal-mining heritage park located in the South Wales valleys. The authors describe how they have constructed an ethnographic hypermedia environment that brings together the presentation of analysed data alongside the accumulated original data. In their paper the authors claim that the strength of the development is its ability to merge different forms of media (visual, verbal and pictorial) thus enabling a deeper ethnographic understanding for readers of their research. The hypermedia environment offers additional dimensions to their presenta- tion of data and the analysis can be made more explicit for the reader to the extent that the reader becomes a co-author exploring and scrutinizing the orig- inal data which are presented through a range of media.
Evaluation
Many of the issues surrounding the evaluation of computer-assisted data analysis focus on questions such as ‘will computers take over the analytical process?’, ‘can computers improve the validity of qualitative analysis?’ and ‘can computers make data analysis more transparent to the readers and users of research?’
Initially many qualitative researchers were suspicious of computer- assisted qualitative data analysis. The computer was considered to be an icon of the quantitative explanatory research paradigm that favoured distance rather than personal engagement with data. More recently computer-assisted data analysis has become more widely accepted due to its advantages in the management and retrieval of extensive amounts of textual data. There are, however, some remaining concerns about the use of computers to assist quali- tative data analysis. Such criticisms include the sacrificing of depth for breadth of analysis and the application of quantitative principles such as frequency counting to qualitative data. Other researchers (see, for example, Coffey, Holbrook and Atkinson, 1996; Seale, 2000) have warned of the simplified asso- ciation of computer-assisted qualitative data analysis with grounded theory. Their concern is that while CAQDAS facilitates the organization and retrieval of qualitative data, software packages are not a substitute for analysis itself. Researchers are still required to undertake the analytic process, to explore the meaning of the data and build theories in their own minds. The software is also considered unsuitable for certain types of qualitative analysis that use Computer-Assisted Data Analysis
relatively short data extracts, such as discourse analysis or conversation analysis, as it has more use in the discovery of the thematic content of qualitative data than the form or structure of talk or text.
A discussion of the perceived merits and disadvantages of computer- assisted data analysis is a regular feature of many research articles which explore qualitative data analysis. It is also a pertinent question for many research students in deciding whether to invest the time to learn the intricacies of a new computer package in order to analyse their data. It is perhaps worth remembering that computer-assisted analysis is no substitute for the analytic mind nor can it offer a shortcut to data coding. Furthermore many of the features that researchers want from computer-assisted analysis, such as search- ing for strings of text, can be found in word-processing packages. There are, however, substantial benefits of CAQDAS particularly in terms of the speed and thoroughness of the data searches and the ability to link data with analy- sis through the use of analytic memos which can aid the transparency of the analytic process.