Qualitative analysis By the end of the next two sessions you should be able to: • evaluate different types of qualitative analyses you read, based on an understanding of their aims and methods • select the type of analysis you would need for a particular purpose, based on an understanding of its methods, advantages and disadvantages • carry out a basic thematic or content analysis of qualitative data Qualitative analysis 1: analysis of meaning Structure of this session Introduction to thematic and content analysis Practice coding Break Introduction to: - Phenomenological analysis - Grounded theory analysis (with practice example) THEMATIC CODING AND CONTENT ANALYSIS Thematic coding refers to any method of categorising segments of qualitative data into meaningful themes Content analysis is a rigorous form of thematic coding that should normally have good inter- rater reliability; codes can then be used as basis for quantitative analysis Available published examples: H.Frith & K.Gleeson (2004). Clothing and embodiment: men managing body image and appearance. Psychology of men and masculinity. 5, 40-48. C.V.Caldicott, K.A.Dunn & R.M.Frankel (2005). Cam patients tell when they are unwanted? "Turfing" in residential training. Patient Education & Counselling, 56, 104-111. Thematic coding Can be used as EITHER a) a means of reducing or summarising the data (‘factual coding’; usually realist and content analysis) OR b) a means of indexing the data (‘referential coding’; usually interpretive and constructivist analysis) In practice, themes serve to identify, label and interpret features of data - simultaneously describe and organise. Thematic coding Thematic coding can be used for any qualitative analytic approach Thematic coding can use any of the procedures of content analysis but: a) reliability less important (as no quantitative analysis used) and b) usually adds more interpretation, based on context of themes identified Content analysis Content analysis is a particular kind of thematic analysis which involves counting instances of particular occurrences, which can be anything of interest: • a) a particular word or phrase, e.g. stress, value for money • b) a semantic (meaning) category, e.g. reference to an event, object, concept – stress = worry, strain etc. • c) a type of utterance, e.g. adjective, verb, laughter, silence Content analysis Interpretation may be based on: • frequency of occurrences (e.g. in different samples, or at different times) • patterns of co-occurrence (e.g. ‘Boolean operators’, cluster analysis) • sequence of occurrences How is coding carried out? Defining coding units Before coding can commence must decide on unit of analysis, which should be ‘most meaningful and productive units that are efficiently and reliably identifiable’ (Krippendorf, 1980) Unit of analysis may be: • entire interview or ‘narrative’ • word, line of text • ‘meaning unit’, ‘sentence’ etc. Developing codes Code may be: • inductive (from data) or deductive (from theory) • manifest or latent Must choose between coding each text segment just once, or allowing segments to have several codes Code development normally involves collaborative process of: • immersion in data, e.g. repeatedly reading transcripts • generating tentative codes • applying and developing codes - refining, elaborating, defining, rejecting, splitting • testing codes for reliability (optional for thematic analysis) Developing a coding manual A good coding manual: a) serves as a ‘paper-trail’ for verification of the analysis; b) makes transparent to the reader the relationship between codes and data; c) makes explicit the processes involved in the construction of the analysis, and the presentation of the interpretation. Coding manual Coding manual should consist of: • label/name for each theme • definition of what theme concerns/name means • description of how to decide when theme occurs • qualifications, elaborations and exclusions • positive and negative examples Determining consistency in coding Can establish consistency/variability in coding: • by same coder on two occasions • by two coders who have collaboratively developed codes • by independent coder using coding manual Calculate agreement by Cohen’s Kappa or by formula: % agreement = number of times both coders agreed _______________________________ number of times coding possible N.B. Rare codes likely to have low reliability for statistical reasons Using SPSS to calculate Cohen’s Kappa: entering data Enter data into SPSS data page with each coder as a variable and each bit of text as a case down the left hand side. If you are calculating the overall Kappa for a complete set of codes give each code a number, and enter the number each coder gave for each bit of text (N.B. SPSS cannot calculate Kappa for more than two raters). Coder 1 Coder 2 1 1 3 3 2 2 2 1 1 1 If you have given some bits of text more than one code you cannot enter multiple values so you must calculate Kappa for each code separately, entering 1 if the code has been coded as present or 0 if it has not. You can also do this if you wish to calculate separate reliabilities for each code (e.g. to see which codes are reliable or unreliable). Using SPSS to calculate Cohen’s Kappa: carrying out the analysis To get SPSS to compute Kappa, click on: Analyse Descriptive Statistics Crosstabs enter one coder as the column and the other as the row Statistics tick the box next to Kappa OK The Kappa value for the sample data I have given should be .69 – you can use this example to check you are doing this right the way. Coding exercise Deductively code text from interview with a woman describing dizziness into Leventhal’s original illness representation categories: 1. Identity (symptoms/characteristics) 2. Cause 3. Timeline 4. Consequences 5. Control/cure Tips for thematic coding and content analysis • The more conceptual the code, the more difficult it may be to agree a definition for and code reliably - but very practical codes may have little analytical significance. • Develop a coding manual as you code each text segment - writing down label, definition, rules for exclusion, good examples of category. First log onto psyweb: www.psychology.soton.ac.uk/psy web Then go to: www.psychology.soton.ac.uk/ psyweb/practicals/ly3/question.php? qn=1 Enter a code for each statement, then submit. www.psychology.soton.ac.uk/quizzes/rm/show_responses.php?qn=1&answer=1 Phenomenological analysis of subjective meaning Origins and aims • Kant (18th century) called for the systematic investigation of ‘phenomena’, i.e. the fundamental, universal contents and organisation of subjective experience. • Heidegger (mid 20th century philosopher) argued we should study the socio-historical evolution of ‘being-in-the-world’, and how this shapes our perceptions. • In the 20th century, humanist phenomenologists have sought to understand the unique experiences and viewpoint of others. Phenomenological analysis - methods • Attempt to discover the meaning of phenomenon through introspection, or empathic identification with interviewee. • Often no formal analytic method, but immersion in data, prolonged reflection, reflexivity. • In Kantian tradition, seek to ‘bracket’ preconceptions, ‘common-sense’ assumptions in order to achieve ‘primitive perception’ or insight into the perspective of the interviewee. • In Heideggerian tradition, consider socio-cultural context, reflexively analyse ways in which meaning co- constructed by investigators and participants • Analysis complete when arrive at rich description and new, consistent, compelling understanding (e.g. Davies, M.L. (1997) Shattered assumptions: time and the experience of long-term HIV positivity. Social Science and Medicine, 44, 561-571.) Phenomenological analysis Advantages • permits imaginative, profound exploration of phenomenon • avoids making claims of ‘objectivity’ • avoids ‘objectifying’/manipulating interviewee • gives ‘voice’ to their experience Limitations • can degenerate into uncritical subjectivity - investigator's or interviewee's - or superficial description • the possibility of ‘primitive’ perception is rejected by constructivists • lack of detailed guidance for analysis can result in inadequate analysis, especially if new to qualitative research Grounded theory Originally (Glaser & Strauss 1960s) claimed to be ‘scientific method’, replacing traditional criteria for reliability/validity with ‘fit, understanding, generality, control’. Features of a comprehensive grounded theory analysis • inductive theory-building (in place of quantification/verification) • iterative process of theoretical sampling and analysis, leading to saturation (i.e. point where collecting further data from new respondents will not yield new information, alter analysis) Techniques of grounded theory • Open and in vivo coding – grounding the analysis in the participants’ words • Constant comparison (between instances of codes in different contexts, between different codes etc.) • Questionning – how, why, where, when, who, with what consequences? • Memoing – creating a paper trail of your thought processes Techniques of grounded theory cont. • Axial coding (create abstract codes from lower level codes, establish their properties and inter-relations) • Diagramming the relationship between codes • Verifying the analysis • Creating a story (often with a timeline) Grounded theory Advantages • explicit, well-disseminated, rigorous methodology • emphasis on inductive method avoids premature imposition on data of pre-existing, imported theory - especially useful for exploratory studies • many very valuable techniques for analysing data Limitations • ‘inductive’ theory-building could be regarded as post hoc from realist perspective, realist from phenomenological perspective • socio-linguistic influences on interviews may not always be well theorised/analysed (but evolving method) First log onto psyweb: www.psychology.soton.ac.uk/psyweb Then go to: www.psychology.soton.ac.uk/ psyweb/practicals/ly3/question.php?qn=1 Write down an ‘in vivo’ code for each statement. What metaphoric connotations does the language participants use have? Then use constant comparison to consider whether statements should have the same or a different code, how codes differ and are connected. Revision questions • What is the difference between inductive and deductive coding? • What are the arguments for and against carrying out inter-rater reliability checks? • What is ‘theoretical sampling’ and why is it used in grounded theory?