Presentation on theme: "EERA Workshop on Q methodology & Q factor analysis"— Presentation transcript:
1EERA Workshop on Q methodology & Q factor analysis Sue RamloJoe JurczykAl HerbertIsadore Newman
2Workshop Outline Introduction to Q methodology Q factor analysis Sorting items onto a normal Gaussian distributionConcourse of items & the Q sampleAnalyzing the sorts using PQ Method – factors & interpretationQ factor analysisHow is Q FA different from Q method?Using QUANAL to find topologies in Q FAUsing PQ Method for Q FAWhy groups of people?
3Introduction to Q methodology - Overview Allows a researcher to investigate the views, or perspectives, of a person or a group of people.The process involves sorting items (typically statements) into a normal Gaussian distribution.Analysis involves factor analysis that groups people, not items. FA – grouping based on correlation among the participants’ sorts.
4Sorting items onto a normal Gaussian distribution Pre-sort into 3 pilesDistribute (& re-distribute) to fit specific normal Gaussian distribution. (usually most disagree to most agree)LEAST like my view(~14 statements here)MOST like my viewNeutral view about this statement
5Now you try it!Condition of instruction - “Because you’ve been working so hard, you have been ordered by your boss/department chairperson to take a 5-day vacation by yourself or with your family, all expenses paid. Rate the following items on a scale of “most enjoyable” (+5) to “most unenjoyable” (-5)….”Pre-sort into 3 pilesDistribute (& re-distribute) to fit specific normal Gaussian distribution. (usually most disagree to most agree)Most unenjoyable(~7 items)MOST enjoyableNeutral view about this statement
6Introduction to Q methodology – historical background Developed in 1935 by William StephensonPhysicist-psychologistStudent of SpearmanA Study of Behavior, 1955Q because of its philosophical similarity to the physics field of Quantum MechanicsMost often used in fields of psychology, marketing, advertising, political science…Mixes quantitative & qualitative aspects of research
7Why not another method to determine views? Alternatives for determining perspectives are not as powerful as Q.Likert scale evaluations and rank ordering lead to the loss of meaning (McKeown, 2001).Because Q measures personal opinion regarding a concourse of items related to a topic, validity is not a consideration (Brown, 1999).
8What is meant by subjectivity? What do you see? Bunny? Duck?Is one right & the other wrong or are they both just different views?
9Any Q study starts with a concourse: Can consist of words / statements, pictures, sounds, smells…SubjectiveNot “It’s raining”But can be “the rain makes me feel sad” –or- “I love to walk in the rain.”Items are interpreted by participant – removes the view of the researcher & the issue of validity / reliability.Select the Q sample from the concourseTry to “balance” the Q sample
10Concourse of items – 3 possibilities “Naturalistic” statements - taken from participants’ oral or written communications.InterviewsFocus Groups“Ready made” statements - from sources other than those of the participants’ communications.Likert survey itemsBased on knowledge of researcher w/o interviewsHybrid - combine both “naturalistic” and “ready made” items.One is not inherently superior to the other (McKeown & Thomas, 1988).Researcher selects the type best suited to the project at hand
11Q sample – select items from the concourse to use in the study. Example: Selection from a Q sample of 44 (chosen from a concourse of 53)1What I learn from textbook depends on how I use it to study in this class.23I can tell when I understand the material in this class.2I need to learn how to study more effectively to succeed in this class.24I feel comfortable applying what I learned in this class to the real-world.3Working hard on difficult problems does not help me learn in this class.25Doing homework helps me learn in this class.4I don't have to work hard to learn in this class.26Reading the textbook helps me learn in this class.5I need to learn how to learn in this class.27I like the exactness of math-type subjects.6I have very little control over how much I learn in this course.28I often think about how well I understand the topics in this class.7In this class, if I can't understand something right away, I will keep on trying29What I learn in this class will help me in other classes.8Working with classmates inside &/or outside this class helps me learn.30What I learn in this class will help me when I get a job in my field.
12Condition of instruction Participants sort based upon a condition of instruction (or multiple conditions).E.g. Sort the following statements as they relate to your views about learning in this class.The statements are matters of subjective opinion and may mean different things to different people.e.g. I worked hard in this class.
13Analyzing Q sortsSPSS & SAS not really designed for Q sorts – you mess with weightings, etc.Need software designed for Q methodologyQUANALPCQPQMethod
14Factor Analysis Higher order correlation Used to determine patterns in a data setR-factor analysis groups items (people are rows, items are in columns). Factors represent similar items.Q-factor analysis groups people (people are in columns, items are in rows). The factors represent people with similar topologies.Q methodology is not Q FA but does group people based upon their VIEWS on a subject. Factors represent similar views about a topic.
15PQ Method to determine factors & assist in their interpretation Free download (start atDOS basedDesigned for handling Q sort entry and analysesChoicesCentroid versus Principle Components factor extractionGraphical hand rotation versus VarimaxStart PQMethod
17Q factor analysis How is Q FA different from Q method? Using QUANAL to find topologies in Q FA
18QFA run in PQMethodCreate z-scores for each item (if the items all have the same scale, skip this step).Rank order items for each person (lowest to highest) and then give a grid score (e.g 1 to 6, -2 to 2, etc). Note that PQ Method allows for a scale up to 13 places.Unlike Q sorts, QFA does not lend itself to a grid that has any columns deeper than one layer. This is important because we are analyzing the topologies of the items.The *.lis file contains similar outputs to that of QUANAL including distinguishing items for pairs of factors, consensus items, etc.
19Why groups of people?If groups are known a priori, use discriminative analysesIf groups are unknown, need Q FA to determine groups without sorts; Q method with sorts