Rachel Baker From understanding perspectives to informing public policy the potential and challenges for Q findings to inform survey design Organised session: Neil McHugh, Job van Exel
Session outline 1. Why Q-to-survey studies? (Baker) 2. Q-to-survey design (McHugh) 3. Five Q-to-survey approaches: did they work? (van Exel)
Session outline (discussion between): 1. Why Q-to-survey studies? (Baker) I. When? Why? Why not? II. III. Some terminology.. Brief history/ non-systematic literature review 2. Q-to-survey design (McHugh) I. Try it! II. What do you think of different approaches? 3. Five Q-to-survey approaches: did they work? (van Exel) I. Discussion of findings and questions
Q-to-Surveys: Before Why, What?! Q studies aim to tell us how people think and feel about and perceive an issue reveal subjectivity Factor loadings give us a quantitative indication of how closely aligned individuals are to factors Surveys measure things like: to what extent a sample of respondents are x or y (how tall, how intelligent) or how many are in categories a, b, or c (male/female; religious/not)
Survey questions based on Q factors Q2S attempting to estimate Q factor association in large respondent samples based on a previous ( standard ) Q study Once the nature of shared views (factors) have been described then it might (on occasion) be interesting to know: How common those views are in a population What the distribution of views are Are some views associated with certain characteristics
Survey questions based on Q factors Q2S attempting to estimate Q factor association in large respondent samples Summarise Q factors in short form By selecting a subset of key statements from the full Q set By describing factors using short descriptions Allow estimation of factor association or counting of factor members
Factor association/ factor membership Factor association Intensity score (akin to factor loading) Tells how much each individual s subjective views is similar to each shared view Factor membership Categorising respondents into factor groups Will likely result in uncategorised or mixed group
A brief history
Stephenson 1953 Study of Behaviour Chapter IX The prior analysis of questionnaires 9
Stephenson 1953 We are to propose that along Q- technique lines it is often possible to discover complex facts, of the kind usually regarded as inferences, by previous study of relatively few cases only. They can thereupon be counted, if need be, by using an appropriate questionnaire and large-sampling techniques p190 10
Talbott 1963 AEJ convention presentation The Q block method of indexing Q typologies 11
12
Brown 2002 Operant Subjectivity paper Selection of statements Likert scale responses 13
Why Q2S?.. and policy Beyond understanding subjectivity and thinking about policy Preserving heterogeneity and still informing decisions Who thinks in what ways? What characteristics are associated with seeing the world differently? Does not necessarily imply majoritarianism? And might still be useful 14
More fundamentally - why not? Do Q survey make reasonable assumptions? By taking the key features of Q factors as flags we can estimate respondents agreement/association with the whole factors Is categorising into factor membership defensible? Or should we stick to intensity scores/ factor association? Does Q2S approach somehow undermine the principles of Q methodology?
Methodological Q2S Research Questions: How well do these and other Q to survey (Q2S) approaches work? Do respondents scores in Q2S approximate their factor association in full Q sorts? (validity) How do different Q2S approaches compare? How reliable/ feasible are they?
Q2S study of 5 approaches MRC funded 3 year study of societal perspectives about the value of life extension at the end of life
GCU factor 3 video https://vimeo.com/productionattic/review/116354872/0c9ac890cf
Surveys estimating factor association and membership Assume 10 survey respondents and 2 factors (red and blue) respondents survey questions 1 2 3 4 5 6 7 8 9 1 0
Surveys estimating factor association and membership Some Issues, illustrated Each respondent has some association with red factor and blue factor. 1 2 3 4 5 6 7 8 9 10
R 0.3 B 0.7 R 0.7 B 0.3 R 0.95 B 0.05 R 0.0 B 1.0 R 0.2 B 0.8 R 1.0 B 0.0 R 0.1 B 0.9 R 0.65 B 0.35 R 0.4 B 0.6 R 1.0 B 0.0 factor association scores 1 2 3 4 5 6 7 8 9 10
Factor membership/ classification nose counting Red Blue Mixed/ uncategorised???? Need rules to categorise and count.. Where to draw the line?
Thank you.