Case Study Research

3 downloads 92710 Views 1MB Size Report
Computer-Based Analysis of Qualitative Data: ATLAS.ti ... abstracted and exported to statistical software such as SPSS. ... software and organized, managed, coded, and ana- lyzed in the ... a tag or label thar best describes the data to which.
- - - - -- --

ENCYCLOPEDIA OF

Case Study Research EDITED BY

Albert J. Mills Saint Marys University, Halifax, Nova Scotia

Gabrielle Durepos St. Francis Xavier University, Antigonish, Nova Scotia

Elden Wiebe The Kings University College, Edmonton, Alberta

volume ($)SAGE los ~es I london I New Delhi Singapore I WashingtOf) DC

A SAGE Reference Publication

182

Computer-Based Analysis of Q ualitative Data: ATLAS.ti

Critical Summary A complex social phenomenon cannot be understood by reducing it to its parts. Rather, a more holistic approach is called for. Complexity is a theoretica l perspective that attempts ro respond to thi s understanding. Case studies viewed through this lens pose a cha llenge to researchers since cases are dynamic and constantly responding to the influences of cultu re and environment. Researchers must be mindful of these influences since they are integra 1 to understanding the cases.

Rosemary C. Reilly and Wanen Linds See also Ecological Perspectives; Webs of Sjgnifican~e

Further Readin gs George, A., & Bennett, A. (2005) . Case studies and theory development in the social sciences. Cambridge: MIT Press. Mainzer, K. (2007) . Thinkillg in complexity (Rev. and enlarged 5th ed .). New York: Springe r. Stacey, R., & Griffin, D . (2005) . A complexity perspective 011 researching organizations. London: Routledge. Stake, R. (1995 ). The art of case study research. Thousand Oaks, CA: Sage. Sumara, D., & Davis, B. (1997). Enlarging the space of the possible: Complexity, complicity and action research practices. Tn T. Carson & D. Sumara (Eds.), Action research as a /iVil1g practice (pp. 299-3 12). New York: Peter Lang. Yin, R. (2003 ). Case study research : Design and methods (3rd ed.). Thousand Oaks: Sage.

COMPUTER-BASED ANALYSIS OF QUALITATIVE DATA: ATLAS.TI ATLAS.ti, which runs on Microsoft Windows, is a software program designed to support the researcher in the interpretation and analvsis of a variety of data sources, including text, audio, and images. Data can be coded, searched, retrieved, codes defined, related codes or documents grouped together, conceptual diagrams of the emerging understanding of the data created, memos written, and tables of numerical data

abstracted and exported to statistica l software such as SPSS. Conceptual Overview and Discussion A case study project is created in ATLAS. ti as a hermeneutic unit (HU) that bundles together all relevant data sources, codes, conceptual linkages, memos, and comments. T he data included in case study research (e.g., text, audio, video, photographs, diagrams, and maps ) are imported into the software and organized, managed, coded, and analyzed in the HU. Data files are referred to as primary documents (PDs). Each PD is numbered in the order in which it is imported into ATLAS.ti-for example, Pi , P2. In case stud y research, where a number of PDs may constitute a case, the numerical ordering of PDs has an important data management function. The PDs for a specific case can be grouped together by importing them concurrentl y. If all PDs fo r a specific case are not imported concu rrently but are scattered among other PDs, the numerical position of the PDs can be eas il y changed to group them as cases. PDs may also be orga nized by cases using a function called "families." A family is a cluster of documents or codes or memos . Within -case and across-case clusters of PDs can also be created. For example, in a case study focused on one or more schools, each school may constitute a case. All data related to a specific school ca n be cl ustered into a family tagged with the name of the school. The PDs of within-case variables such as teachers and students can be clustered in PD famil ies labeled "Teachers" and "S tudents," respectively, separating the PDs according to cases and "ariables wirh in cases. Across-case groupings can be similarly clustered. Grouping PDs in families facilitates a focus on a specific case and/or a specific variable set, unencumbered by the presence of PDs that do not belong to that case or variable. In ATLAS .ti, data mUSt be coded to access fu rther functions of the software. Cod ing refers to the assigning of a PD to one or more codes. A code is a tag or label thar best describes the data to which it is assigned. A margin display alongside a text- or image-based PD provides a visual cue of coding as it occurs. There is no margin display for video or audio PDs.

Computer-Based Analysis of Qualitative Data: ATLAS.ti

There are at least two types of codes that organize the data: data management codes are mutually exclusive, and describe the characteristics of a PD; for example, "adolescent" or "female. " Characteristics of a PD are often sociodemographic variables. The whole PD is assigned to the relevant data management codes. Conceptual codes assign meaning drawn from the data (inductive) or from theory (deductive), and are generally not mutually exclusive. Segments of a PD (e.g., lines of text, portions of a photograph, seconds of an audio file, or frames of a video) are assigned to multiple conceptual codes. Codes are organized in a code pane. The "groundedness" of a code refers to the amount of data coded to a specific code. How grounded a code is in the data is indicated by a number immediately next to the code in the code pane, indicating how many times the code has been assigned in the data. A table for all codes illustrating their gro undedness can be exported from the software to Excel or SPSS. A definition for each code, to increase coder reliability, can be inserted in the blank comment space beneath the code in the code pane. Searching for words in text PDs can assist the coding process by quickly finding specific words that may signify a code. Automatic coding makes use of the word search function to assign words-incontext to one or more codes automatically. The context (a word, sentence, paragraph, multiple paragraphs, or the whole document) can be selected when "scoping " the search. Word search and automatic coding functions cannot be used with nontext PDs such as photographs, maps, aud io, and video files. As coding progresses, relationships among the codes may become apparent. Codes that are related may be grouped together in code families, with each family designating a theme or category. By selecting a specific code family, access is provided to only the codes in that theme or category. _-\nention can be focused on the codes that constitute the code family without being distracted by other codes. An in-depth, focused analysis of a case by a theme can be facilitated by selecting the PD family that represents the case (excluding all other PDs) and the code fam ily representing a theme (excluding all other codes and themes). Code and PD families can be used to delimit or scope a search using the query tool in ATLAS.ti. To make the best use of the guery tool for searches,

183

the PDs must be coded. Assigning PDs to both data management and conceptual codes facilitates the retrieval of data. The clustering of PDs into fam ilies allows searches to be limited to a case, as well as to within- and across-case searches. The results of a search are reliant on the quality of coding of PDs. Boolean searches combine, intersect, and subtract coded data . Proximity searches show relationships between coded data such as whether data overlap, enclose, or are near to other coded data. Search and retrieval enables further in-depth exploration of the data. Writing is an important aspect of case study analysis, and within ATLAS.ti, comments and notes may be written in the form of memos, providing a chronological trai l of developing thoughts and concepts. By writing within the software, new comparisons, insights, themes, and relationships between codes may emerge, deepening initial understandings and explanations. Two types of display in ATLAS.ti that are useful for case study research are matrices or tables, and semantic networks. Matrices of codes, showing coding as number of words or number of times a section of a PD has been assigned to a code, can be exported from ATLAS.ti to a statistical package. The network view tool allows the researcher to ass ign relationships between codes. Codes and memos are easil y imported or created in the network view. Causal, associative, and other relationships between codes can be assigned. New relation ship types can be created in the Relationship Editor. An iterative and interactive process of working between network views and the analytic text of memos can lead to the discovery of further relationships, understand ings, and explanations. A hypertext function links selected sections of one PD to selected sections of the same PD and/or other PDs. These links can be traversed by following the symbols denoting hypertext links in the display margin, or can be visually displayed in the network view. Causal , associative, and other relationships between hyperlinked segments can be assigned. Data can be exported from ATLAS.ti. Output including the text assigned to one or more codes, references to nontext data, PD lists, codes and their definitions, and memos can be brought to screen, sent to file, or printed.

184

Computer-Based Analysis of Qualitative Data: CAITA (Computer-Assisted Interpretive Textual Analysis)

Application Vaishali Patel and Anne Riley used ATLAS .ti to organize, manage, and analyze data in a multiple case study examining use of Outcomes Management System (OMS) data in decision making by staff in out-of-home childcare programs. Type of staff and service setting were defined as cases, and crosscase comparison of OMS data usage was undertaken. PDs were grouped according to case and, following coding in the software, reports displaying selections of text assigned to specific codes were generated and read to further the fragmenting of codes. Memos documented emerging patterns, meanings, and understandings . Additional reports of text coded to two or more codes were generated and patterns in the data identified, compared, integrated, and summarized. The authors' reported an iterative and cyclical process of coding, memo development, repons, and further reading of PDs to deepen the case study analysis. Critical Summary ATLAS .ti is a storage, organizational, management, and analysis tool for case study research. A variety of data types can be imported into ATLAS.ti, entiching the study. PDs can be grouped together in cases, and memos and codes grouped thematically, facilitating within- and across-case analysis . Codes are not linked to any other code until the researcher specifically creates the relationship, differing from many software where hierarchical relationships between codes are encouraged . Codes and themes can be easily interrogated in ATLAS.ti, enhancing analysis and the construction of theory . B. Raewyn Bassett See also Inductivism; Iterative; Reliability; Visual Research Methods; \XTirhin-Case Analysis

Further Readings Lewins, A., & Silver, C. (2007) . Using software in qualitative research: A step-by-step guide. London: Sage. Patel, V. N ., & Riley, A. W. (2007). Linking data to decision-making: Applying qualitative data analysis methods and software to identify mechanisms for using outcomes data. Journal of Behavioural Health Services & Research. 34, 459-474 .

COMPUTER-BASED ANALYSIS OF QUALITATIVE DATA: CAITA (COMPUTER-AsSISTED INTERPRETIVE TEXTUAL ANALYSIS) Computer-assisted qualitative data analysis nses computer hardware and software to support qualitative data analysis tasks, including search and recovery of data, representation of data, summarizing and interpretation of themes, and exploration of meanings and patterns found in data . Qualitative data are primarily textual data, for example, newspaper articles, organizational documents, interview transcripts, and narratives. Computer-based analysis of qualitative data can involve two forms of analysis: Quantitative analysis of qualitative data uses computers to count key words or codes and is called computer-based content analysis. Qualitative analysis of qualitative data uses nonmathematical processes to explore data. One specific approach to qualitative analysis of qualitative data is computer-assisted interpretive textual analysis, which explores members' meanings in qualitative data through theoretical sampling, computer software, and expansion an alysis of data displays. Conceptual Overview and Discussion Quantitative analysis of qualitative data uses computer technology to quantify qualitative data and to perform content analysis. Computer-based content analysis of qualitative data involves breaking textual data down into segments, then coding or classify ing each segment. Researchers then use computer technology to count key words or codes and to compose operational indicators of variables. The quantitative measures are used to test or evaluate hypotheses. This reflects a positivistic paradigm where quantitative variables are created and relations among variables are used in a deductive manner to test or confirm theory. The second form of computer-based analysis of qualitative data uses qualitative analysis involving nonmathematical processes of interpretation to understand the lneanings or patterns in qualitative data . The central task of qualitative analysis is to understand the meaning of the text (i .e., qualitative