Analysing Qualitative data

Chapter 9 CMI

Analysis of Qualitative data

Assumed researcher can analyse and interpret the data as they are collected. While data is being gathered, researcher often more concerend with process of the research than details of the content. Interviewer bias may occur, so content recalled is content that supports researcher or clients point of view, or comments of particular vociferous or articulate respondetns are recalled more easily than others.

Taking stock at the end of the project is prefereable as the researcher can be reimmersed in all of the data gathered and can organise the content into a form that directly answers the research objectives.

Content analysis

the analysis of any form of communication, whether its advertisements, newspaper articles, television programmes or taped conversations. Involes 2 main components:

1. Organisation of the data: structuring, ordering of data using manual or computerised procedures

2. Interpretation of the data: determining what the data say with regard to the research objectives 

Transcripts

Detailed ‘word-for'-word’ records of the depth interview or group discussion setting out the questions, probes and participant answers. 

Organising the data 

Researchers generally obtain transcripts of interviews or dicussions to work from as its hard to analyse segments of different interviews whilst having to move back and forth between different sections of recordings, dont discard the videos. Help to determine strengths of respondents feelings through highlighting the expressions and tone of voice used when certain points are made. Significant care should be taken in preparation.

Coding (reducing)

the procedures involved in translating reponses into a form that is ready for analysis. Normally involves assigning codes to responses. Processing the data into discrete chunks and creating a reference and categorisation mechanism for these chunks of data so they can easily be retrieved for analysis. Codes are labels that assign meaning to the data collected, done in a number of ways.

Time-consuming process but forces researcher to focus upon the key elements appearing in the data. To understand meaning of these elements and their interrelationships, neccesary to display the elements of the qualitative data in an accessible and abbriviated format

Data display

Summarising and presenting of information in order that relationships or connections can be identified and conclusions can be drawn.

Also allows others to view the logic behind the manner in which the relationships and conclusions have been reached. The most common manual approaches are: the tabular/cut-and-paste method and spider-type diagrams

Tabular/cut-and-paste method of analysis 

A method for analysing qualitative research data using a large sheet of paper o an Excel sheet divided into boxes. Use 2 or 3 most important respondent characteristics being put into column headings. The row headings are the most important issues relating to the research objectives for the project, researcher then transfers content into relevant boxes. When interviews are structured in this manner, it becomes easier to compare and contrast the data collected. 

Ensures that each transcript is treated in the same way, therefore allows more researchers to undertake the task simultaneously. It’s important as it takes a lot of time to go through each transcript, particularly when it relates to a group discussion of up to 2 hrs. 

Disadvantage is its inflexible, info that doesnt fit the framework is ignored, even if valuable in another manner to the research and client. Can be labourious exercise copying over material into grid of boxes 

Spider-type diagrams (mindmapping)

Diagrams used to organise data in the analysis of qualitative research data. Sets out each of the issues at the centre of the diagram with key responses emanating from the centre like a spiders web, highlighting links between different issues. Interconnections are much clearer to the researcher and client than the case in tabular format. Often difficult for more than one researcher to undertake the task simultaneously as unlikely to categorise the linkages in the same manner.

Data validation

The verification of the appropriateness of the explanations and interpretations drawn from qualitative data analysis

Researcher needs to demonstrate that the explanations and conclusions drawn from qualitative data are not just a reflection of their own views as own opinions could unintentionally influence how the data is interpreted. 3 most common validation approaches are triangulation, inter-coder reliability and participant validation.

Triangulation 

Using a combo of different sources of data where the weaknesses in some sources are counterbalanced with the strength of others. Minimum of 3 reference points are taken to check and objects location. Comparing info from different kinds of data and different methods allows researcher to seek corroboration for the qualitative research findings. May consider using published data or undertaking a programme of quantitative research or even observations and photographs to validate qualitative findings. 

Inter-coder reliability

An approach aimed at assessing the reliability of coding by comparing the coding of an independent coder with that of the original researcher for the same body of data. Independent coder should have familiarity with the broad subject area of the research and undertake coding without viewing the coding created by the original researcher. Only tested on a proportion of transcripts than the whole data, depends on number of transcripts avaliable.

Realiability assessed on extent which coding completed by the independent coder deviates from 100% agreement with original researcher. Accepted reliability standards vary but generally would be concerns about objectivity of the researchers coding if the lvel of agreement was less than 90%. Coding differs then may need to look at particular codes and application to data.

Participant validation

This involves taking the findings from the qualitative research back to the participants/respondents who were involved in the study and seeking their feedback.

If feedback verifies the explanations and conclusions, then researcher can be confident about validity of the findings. More commonly used in research involving b2b markets or academic research projects

Qualitative analysis software

programs can also be used to categorise and order the data. Computers assist with sheer volume of data collected. 2 main broad groups of programs are content analysis software and text analysis/theory-building software.

Content Analysis software

Software used for qualitative research that counts the number of times that pre-specified words or phrases appear in text. 

Can do various operations with what it finds, like marking or sorting text found in new files and production of text frequency distributions - set out number of times certain texts appear in transcripts, provide initial comparisons between transcripts to determine main topics or themes that they share. Analysis of this type is crude and only value giving broad first feel of data rather than form of detailed analysis. 

Text analysis software

Software used for data analysis in qualitative research. Tey help segment data (attach a code) and identify any patterns that exist.

Researcher can then retrieve data where one category or topic is dicussed in relation to another. Patterns and relationships within the data may then be identified through using the softwares system of rules and hypothesis-testing features, frequently based on formal and Boolean searcing logic. Some programs search for overlapping or nested segments. Software allows the researcher to build up tree displays of categoeis that can continually be altered as the researcher refines the categories and relationships between them.

Factors when selecting software for qualitative anaylysis

1. Data entry: Program allow researcher to format text in different ways or follow strict rules relating to numbers of characters per line and delineation of blocks of text?

2. Data storage: data be stored in original word-processed format or split into separate files, harder to change?

3. Coding: coding done on screen or work with hard copy to assign codes before entering into the computer? how complex the coding? allow multiple codes or single block? allow making marginal notes or annotations?

4. Search and retrieval: how fast the package search for and retrieve appropirate strings of data? what info is displayed on result of searches? does it make record of searches done?

Advantages of qualitative analysis software 

• Can sort and search through large amounts of data, gives more time to think about what the data means than focus on operational tasks managing the material

• Can Share the data with other researchers, easier to undertake measures of inter-coder reliability while allowing for discussions about findings that emerge from data.

• Can provide sense of rigour, using software means themes in data less likely to be missed, provide greater objectivity of measurement about scale of agreement with certain attitudes and opinions 

Disadvanatges of qualitative analysis software

• Intricacies of the software packages: takes time to learn, especially with including pictures, notes and observations alongisde the transcript of an interview or group discussion. Researchers working to client-imposed time pressures could be unwilling to allocate the effort needed to fully utilise such software.

• Fragmentation of data: software breaks data down into chunks but sometimes researcher needs complete transcripts to fully understand whats happening and attitudes of people.

• A mechanistic approach: software cant understand way words are used, researchers can understand relationship between words that may suggest sarcasm or sentiment different from one identified by software.

Interpreting the data

Computer programs are only capable of organising the data but not interpret meaning. Researcher must interpret and determine what it means, identifying the implications for marketing decisions. Must think through the manner in which things were said to determine what messages the respondents were trying to communicate, conciously or subconsciously. Researchers understanding of this increases with experience of qualitative research but from personal expereinece or socially interactive situations in everyday life. We interpret commitment, belief and enthusiasm, noticing inconsistencies, uncertainty, fear and incomprehension. Can recognise boasting, bluffing and disguising that people do to cover true opinions and behaviours. Combining intuitive skills with understanding literature of group dynamics can help understand respondents. Helps determine the implications for the research objectives for marketing decisions to be made.

Grounded theory: An academic approach 

A systematic method of generating theory and understanding through qualitative data collection and analysis. More commonly used by academic researchers rather than marketing researchers in areas where little is known about the subject or where new approach to understanding behaviour is required. 

Unlike qualitative research where data is collected then analysed, researchers start with a very open agenda, collecting data and analysing them simultaneously from the outset. 

They start with a broad idea of interest, then use preliminary interviewing to open up this area. Examine the initial info and participant views to develop further questions around these views and seek out further participants who could potentially elaborate on the topics/views identified. Repeated many times as researcher refines the emerging themes and conclusions as they go along. Researcher therefore, needs to be involved at  each stage of data collection and analysis

Data analysis ongoing throughout data collecting period and afterwards have 4 components

1. Coding data: more detailed and guides further data-gathering as project proceeds. Starts with open coding, line-by-line basis, everything coded to understand issues and concepts. Then axial coding develops overarching topic categories and exploration of relationships between these categories - identified by data and not yet been determined from past research or other theories. Consideration given to the causes, actions and interactions of this axial code with other axial codes identified. Selective coding stage, topic categories and categorial interrelationhsips are used to create a stroyline that tells or explaisn the phenomenon.

2. Memo-writing: Throughout the process, grounded theorists write memos to elaborate on their codes by identifying arguments/ideas that come about as they go through the data. More detailed memos means raising the theoretical level of their work. Process produces ides on concepts, meanings and ideas on whats emerging from the data, also gaps and missed opportunities in early data collection.

3. Theoretical sampling: further data-gathering driven by ideas and concepts emerging from coding and memo-writing. Aim to seek out other situations or participants that may assist in discovering variations in concepts and theories identified. Thinking about concepts should become more precise with further research inputs.

4. Integrating analysis and sorting: various memos are sorted, result in new ideas emerging and recorded in further memos. Emphasis on connections with focus on how ideas fit together. Rssults in a collection of explanations that explains subject of the research.

Pros/cons of Grounded theory

provides significant depth of understanding of a topic

but very time-consuming, difficult to predict length of the study.

Good for academic research but inappropriate for marketing research projects that meet client-determined timescales.