Assessment |
Biopsychology |
Comparative |
Cognitive |
Developmental |
Language |
Individual differences |
Personality |
Philosophy |
Social |
Methods |
Statistics |
Clinical |
Educational |
Industrial |
Professional items |
World psychology |
Statistics: Scientific method · Research methods · Experimental design · Undergraduate statistics courses · Statistical tests · Game theory · Decision theory
Q Methodology is a research method used in psychology and other social sciences to study people's "subjectivity" -- that is, their viewpoint. Q was developed by psychologist William Stephenson. It has been used both in clinical settings for assessing patients, as well as in research settings to examine how people think about a topic.
The name "Q" comes from the form of factor analysis that is used to analyze the data. Normal factor analysis, called "R method," involves finding correlations between variables (say, height and age) across a sample of subjects. Q, on the other hand, looks for correlations between subjects across a sample of variables. Q factor analysis reduces the many individual viewpoints of the subjects down to a few "factors," which represent shared ways of thinking. It is sometimes said that Q factor analysis is R factor analysis with the data table turned sideways. While helpful as a heuristic for understanding Q, this explanation may be misleading, as most Q methodologists argue that for mathematical reasons no one data matrix would be suitable for analysis with both Q and R.
The data for Q factor analysis comes from a series of "Q sorts" performed by one or more subjects. A Q sort is a ranking of variables -- typically presented as statements printed on small cards -- according to some "condition of instruction." For example, in a Q study of people's views of George W. Bush, a subject might be given statements like "He is a deeply religious man" and "He is a liar," and asked to sort them "from most like how I think about George W. Bush, to least like how I think about George W. Bush." The use of ranking, rather than asking subjects to rate their agreement with statements individually, is meant to capture the idea that people think about ideas in relation to other ideas, rather than in isolation.
The sample of statements for a Q sort is drawn from a "concourse" -- the sum of all things people say or think about the issue being investigated. Since concourses do not have clear membership lists (as would be the case in the population of subjects), statements cannot be drawn randomly. Commonly Q methodologists use a structured sampling approach in order to ensure that they include the full breadth of the concourse.
One salient difference between Q and other social science research methodologies, such as surveys, is that it typically uses many fewer subjects. This can be a strength, as Q is sometimes used with a single subject. In such cases, a person will rank the same set of statements under different conditions of instruction. For example, someone might be given a set of statements about personality traits and then asked to rank them according to how well they describe herself, her ideal self, her father, her mother, etc.
External links[]
- Q Methodology page Includes more information on Q, as well as free software for conducting a Q factor analysis.
Q Methodology
The Q Method Page (International Society for the Scientific Study of Subjectivity)
Q Methodology takes an inventory of an individual's subjective positions on issues. It aims not to identify mean or average values, but to identify the spectrum of positions. It provides a number of potential uses:
analysis of contributions to dialogues audit the effects of deliberation/participation on social learning filter candidates for delibertive processes for issue representativeness The steps:
Concourse definition is identifing the full breadth of social discussions and discourses surrounding the problem or issue: i.e., the "concourse." The extent of the research is limited only by the constraints of resources and time. Everything from newspaper articles and PR advertising to political speeches and neighbourhood discussions are legitimate sources. Reading, interviews and surveys can all be of value.
Q-set selection is based on the large number, possibly hundreds, of statements on the topic distilled from the concourse. This original sampling is then distilled to a more manageable number, usually no more than sixty, proportionately representative statements: the Q set.
P-sample is the set of participants in the relevant process. Usually the P-sample involves no more than fifty.
Q-sorting involves all members of the P-sample rating the statements in the Q-set on a Likert scale. The distinctive element of Q-sorting is that all statements are ranked and scored together, generating for each respondent a complete partial ordering of all statements scored across the spectrum
Statistical analysis subjects Q-sorts to factor analysis, enabling identification of clusters of Q-sorts containing ranking patterns.
Result interpretation requires careful examination of the rankings assigned to Q-statements by members of each cluster. Some times unanticipated elements emerge that may require researchers to reassess previous assumptions.
Toddi A. Steelman, Understanding Participant Perspectives: Q-Methodology in National Forest Management (online paper)
Steven R. Brown, The History and Principles of Q Methodology in Psychology and the Social Sciences (online paper)
Q-Methodology: A Primer The following information consists of postings in late 1991/early 1992 to QUALRS-L@UGA, a qualitative methods list out of the University of Georgia. A revised version was published as AA Primer on Q Methodology, Operant Subjectivity, 1993, 16, 91-138, by Steven R. Brown.
Q METHODOLOGY Introduction
In the latter half of September, as the Qualitative Research for the Human Sciences list (QUALRS-L) was just beginning, mention was made of Q methodology and its connection to qualitative research methods. (The first mention was on September 22, by Michael Foley, who was responding to someone's inquiry concerning the use of correlation in discourse analysis.) The discussion quickly died, but was renewed in late October when Robert Mrtek recommended Q to a quantitative analyst (as the person characterized himself) who was interested in combining qualitative and quantitative methods. Jean Latting then asked if there was any step-by-step information about Q technique, to which Mrtek then responded with a list of references and with brief mention of the use of SPSS for data analysis. The discussion was then joined by Arthur Kendall, Rich Hofmann, and myself, and in some instances theoretical and conceptual disagreements were apparent. Subsequently, exchanges on the list were supplemented by private correspondence, one of which came to me as follows:
... maybe for the rest of the list you could explain, in simple terms, exactly what Q methods are good for -- in other words, what are they going to tell me about a phenomenon that I cannot learn some other way?
This I propose to do in a series of short notes -- the first of which is appended -- designed to provide a basic understanding of the main features of the methodology. Due largely to its mathematical substructure, Q is fairly well known in quantitative circles, but it is hoped that the following introduction will alert subscribers to this list of its significance for qualitative research as well, including what it might "tell me about a phenomenon that I cannot learn some other way."
1. Background
What is currently referred to as "Q methodology" was introduced by psychologist/physicist William Stephenson (1902-1989) in a letter to Nature in 1935, and spelled out in more detail in "Correlating Persons Instead of Tests" (1935), "Foundations of Psychometry: Four Factor Systems" (1936), and in a celebrated paper with Sir Cyril Burt ("Alternative Views on Correlations Between Persons," 1939) in which the two laid out their contrasting views. His major statement is The Study of Behavior: Q-technique and Its Methodology (1953).
In large measure, the differences of opinion which have recently appeared on QUALRS-L can be traced to the theoretical divergences of the 1930s. Burt's viewpoint, bolstered by such notable factor analysts as R.B. Cattell, Hans Eysenck, and L.L. Thurstone, has generally carried the day and has been ensconced in research methods texts in a variety of fields, not to mention users' manuals for SPSS and other statistical packages, which helps explain why Stephenson's views often sound so out of step despite the fact that Q methodology was his innovation.
Recently, however, Stephenson's ideas have gained in prominence outside psychology. Spurred initially by his own The Play Theory of Mass Communication (1967/1988), a number of other books and articles have appeared which have served to clarify Q's presuppositions and to demonstrate its applicability in virtually every corner of human endeavor. In 1977, publication began of Operant Subjectivity: the Q Methodology Newsletter (now in volume 14), which was recently adopted as the official journal of the newly created International Society for the Scientific Study of Subjectivity. The Society has met annually since 1985 and has generally pursued the implications and applicability of Stephenson's ideas in psychology, communication, political science, health, environmental, and related areas. On-going exchanges are also to be found on QTemp, a Bitnet list accessible via ListServ@KentVM. (Webmaster's Note: These exchanges are still be carried on now at the Q-Method ListServ.)
Fundamentally, Q methodology provides a foundation for the systematic study of subjectivity, and it is this central feature which recommends it to persons interested in qualitative aspects of human behavior. Most typically, a person is presented with a set of statements about some topic, and is asked to rank-order them (usually from "agree" to "disagree"), an operation referred to as "Q sorting." The statements are matters of opinion only (not fact), and the fact that the Q sorter is ranking the statements from his or her own point of view is what brings subjectivity into the picture. There is obviously no right or wrong way to provide "my point of view" about anything -- health care, the Clarence Thomas nomination, the reasons why people commit suicide, why Cleveland can't field a decent baseball team, or anything else. Yet the rankings are subject to factor analysis, and the resulting factors, inasmuch as they have arisen from individual subjectivities, indicate segments of subjectivity which exist. And since the interest of Q methodology is in the nature of the segments and the extent to which they are similar or dissimilar, the issue of large numbers, so fundamental to most social research, is rendered relatively unimportant. In principle as well as practice, single cases can be the focus of significant research.
In short, the focus is all on quality rather than quantity, and yet some of the most powerful statistical mechanics are in the background, but sufficiently so as to go relatively unnoticed by those users of Q who are disinterested in its mathematical substructure. What this might mean for the student of qualitative methods is illustrated in a single study, which will be serialized in the days to follow.
2. Concourse Theory
As noted in Part 1 (Background), Q methodology is comprised of procedures and a conceptual framework that provide the basis for a science of subjectivity, and its phenomena consist of the ordinary conversation, commentary, and discourse of everyday life -- of the kind that proliferates, for example, when discussion turns to such things as the Gulf War, the care of geraniums, whether we can trust Boris Yeltsin, pornography, literary and popular impressions about the movie The Silence of the Lambs, psychotherapeutic strategy, the meaning of life, what to do about the current recession, and so forth.
In Q, the flow of communicability surrounding any topic is referred to as a "concourse" (from the Latin "concursus," meaning "a running together," as when ideas run together in thought), and it is from this concourse that a sample of statements is subsequently drawn for administration in a Q sort. The best references on concourse theory are William Stephenson's "Concourse Theory of Communication" (1978), "Consciring: A General Theory for Subjective Communicability" (1980), and "Protoconcursus: The Concourse Theory of Communication" (1986).
Concourse is the very stuff of life, from the playful banter of lovers or chums to the heady discussions of philosophers and scientists to the private thoughts found in dreams and diaries. From concourse, new meanings arise, bright ideas are hatched, and discoveries are made: it is the wellspring of creativity and identity formation in individuals, groups, organizations, and nations, and it is Q methodology's task to reveal the inherent structure of a concourse -- the vectors of thought that sustain it and which, in turn, are sustained by it.
By the same token, concourses are not restricted to words, but might include collections of paintings, pieces of art, photographs, and even musical selections. In his dissertation on "The Shifting Sensorium" (1990), for example, Paul Grosswiler has created a multimedia Q sort comprised of writings, snippets from videos and records, and pictures; and in his recent paper on "Humor Communicability" (1991), Dennis Kinsey employs as "statements" a selection of Gary Larson cartoons. The idea of concourse incorporates virtually all manifestations of human life, as expressed in the lingua franca of shared culture.
A concourse can be gotten in a number of ways. The most typical is by interviewing people and jotting down or recording what they say, but commentaries from newspapers, talk shows, and essays have also been used. The level of discourse dictates the sophistication of the concourse: hence, factors which should be taken into account in decisions about who should receive a liver transplant at a particular hospital would likely involve the medical personnel, the potential recipients (and perhaps the donor), and possibly even a philosopher specializing in medical ethics (or sociologist with expertise in medical sociology) who might be called in as a consultant. A study of public opinion, on the other hand, would necessitate interviewing representatives of those segments of the society apt to have something to say about the issue in question.
An illustration might be useful at this point to give substance to the above generalities, and for convenience we can take the commentary that was generated on QUALRS-L about Q methodology itself -- from about September 22nd until early this month. Readers unfamiliar with Q methodology will not be surprised to find that much of the commentary to follow is of a specialized nature, hence comprehendable in detail by a relatively small audience; the same could be said, however, of a qualitative analysis of clients in therapy or members of a delinquent gang: a subculture has specific issues which are central to it, and often a specialized language evolves for expressing ideas that may appear obscure to the outsider (who nevertheless may see things more clearly by virtue of being outside). What follows are just a few of the elements from the small concourse which was generated. The names of authors of the comments follow in parentheses, along with the date of the QUALRS-L transmission:
* It allows us to sort patterns of speech among speakers. (Michael Foley, 9/22) * It uses an ipsative technique of sorting a representative set of subjective statements drawn from a concourse of possible feelings or reactions about a subjective condition. (Robert Mrtek, 10/27) * In Q-factor techniques, a case by case matrix of some sort of similarity measure (usually an ipsatized correlation) is analyzed. (Arthur Kendall, 10/28) * Q factor analysis is a simple variation of factor analysis, actually component analysis. (Rich Hofmann, 10/28) * Q methodology is a set of procedures, theory, and philosophy supporting the study of the same kind of subjectivity that is the focal point of much qualitative research. (Brown 11/4)
(The original commentary from which the above were abstracted was naturally more detailed. A copy of the original can be gotten by sending a private request for the file QUAL CONCOURSE to SBrown@KentVM.)
As is apparent, the statements in the concourse are subjective as opposed, say, to the statement that "correlation is a statistical method," which is uncontroversial and ostensibly true. Concourses such as the above comprise the raw material of a human science in its subjective respects, and it is frequently at this point that a qualitative analysis breaks down. Once "texts" (in the widest sense) have been gathered -- from interviews, diaries, participant observation, etc. -- the task becomes one of organization, analysis, and presentation, and in most instances the observer is forced to fall back (as in content analysis) on categories which are superimposed on the data. As will be seen in the sequel, Q methodology likewise involves the artificial categorizing of statements, but ultimately this artificiality is replaced by categories that are operant, i.e., that represent functional as opposed to merely logical distinctions.
3. Q Samples
In Part 2, it was noted that concourse comprises the raw materials for Q methodology, and for the human sciences generally insofar as they are concerned with life as it is lived, i.e., from the vantage point of the person involved. The example rendered consisted of the brief commentary that had accrued on QUALRS-L concerning the nature and scope of Q methodology itself. Diversity in viewpoint was abundant, from the technicalities of factor analysis to the abstractions of quantum theory, from the simplicity of Q sorting to more complex philosophical considerations about subjectivity. The concourse is far from complete and could, if desired, be supplemented with comment and controversy dating from the mid-1930s, when Q was born. Still, what has appeared on QUALRS-L is sufficiently comprehensive to demonstrate a range of opinion, and to introduce the problem of what to do with all the assertions that have been entered into the discursive arena.
For experimental purposes, a subset of statements, called a "Q sample," is drawn from the larger concourse, and it is this set of statements which is eventually presented to participants in the form of a Q sort. The statements selected for this particular study are as follows:
Q SAMPLE FOR QUALRS-L PERSPECTIVES ON Q (1) It permits the a priori structuring of hypotheses in the design of the Q set to be sorted. (2) Q methodology is a set of procedures, theory, and philosophy supporting the study of the same kind of subjectivity that is the focal point of much qualitative research. (3) The method can be coupled with analysis of variance to test hypotheses. (4) The interpretation of factors is more difficult if the Q sorts are internally inconsistent than when they are based on structured Q sets representing testable scientific hypotheses. (5) Centroid factor analysis is recommended since its indeterminacy is compatible with quantum theory and, at the rotational stage, with interbehavioral principles. (6) "Ipsative" generally applies to patterns of objective scores for persons, and has little to do with the subjectivity intrinsic to Q methodology. (7) Cluster analysis may bear some statistical similarity to Q factor analysis, but in most respects it is quite different from the version of factor analysis used in Q methodology. (8) The history of Q methodology attests to the largely arbitrary division between qualitative and quantitative. (9) Cluster analysis is really something quite different and has no commitment to that subjectivity which is central to Q methodology. (10) Variance designs are only used to represent theory. Testing is in terms of dependency factor analysis. (11) The idea is to come up with a set of traits that characterize individuals, then compare individuals for the distribution of these sets. (12) Q can give some fascinating insight into underlying philosophic structures which comprise subjective phenomena. (13) It is intended to get at patterning within individuals (case-wise) rather than simply across individuals (factor-wise sorting). (14) It allows for the interpretive study of subjective behaviors without imposing the usual biases of structured survey questionnaires. (15) Q-factor is an early form of cluster analysis. (16) Factor scores can be tough to come by because the correlations are of reduced rank. (17) There is more to the method than just the technique of Q sorting. (18) Q has never involved the correlation and factor analysis by rows of the same matrix of data that is analyzed by columns in R methodology. (19) The frequencies in the piles must be restricted to the frequencies that would be expected if you had a normal curve, with each pile corresponding to an area of a normal curve. (20) It uses an ipsative technique of sorting a representative set of subjective statements drawn from a concourse of possible feelings or reactions about a subjective condition.
As with sampling persons in survey research, the main goal in selecting a Q sample is to provide a miniature which, in major respects, contains the comprehensiveness of the larger process being modeled. The problem, of course, is how to select from the concourse so as to provide representativeness in the Q sample, and the main device relied upon to achieve this is Fisher's experimental design principles (see Brown, "On the Use of Variance Designs in Q Methodology" (1970)).
In this particular case, the simplest of designs was employed. While perusing the concourse, it was noted that some of the statements were of a technical nature, viz.:
The method can be coupled with analysis of variance to test hypotheses.
On the other hand, there were comments of a more abstract and methodological nature (methodological, that is, in its wider and more philosophical sense):
Q can give some fascinating insight into underlying philosophic structures which comprise subjective phenomena.
As a preliminary matter, therefore, all statements in the concourse were categorized as either (a) methodological or (b) technical, depending on their main thrust, all the time recognizing that few statements are ever one or the other exclusively.
It is often the case that more than one dimension (e.g., methodological/technical) is at issue, and so at this point we could have subdivided the (a) and (b) statements above -- e.g., into (c) Stephenson, (d) Burt, and (e) Neither, to take into account the intellectual heritage of the points of view at issue. This would have provided the following design, with 2x3=6 cells:
(a) methodological (b) technical
(c) Stephenson (ac) (bc) (d) Burt (ad) (bd) (e) Neither (ae) (be)
Equal numbers of statements would then be selected from each of the cells (e.g., 8 of type ac statements, 8 of type ad, etc.) for a Q-sample size of N=(6)(8)=48 statements for Q sorting by respondents.
To keep matters simple, only N=20 statements were chosen for this illustration, 10 from category (a) methodological and 10 from (b) technical. The statements in each category are as follows (numbers are associated with the above statements): (a) Methodological: 2 5 6 8 9 12 14 17 18 20 (b) Technical: 1 3 4 7 10 11 13 15 16 19
As can be seen, the statements are numbered randomly. They are then typed one to a card, much as appears above. The result is a pack of cards (numbering 20) ready for Q sorting.
Before concluding this section, it is important to note that, unlike scaling theory, no assumption is made that the 20 statements above in any sense measure a "methodological" or "technical" position or stance or understanding per se. In The Study of Behavior (1953, chap. 2), Stephenson distinguishes among general, singular, and induced propositions, and the a priori placing of statements into this or that category is exemplary of the former: a statement can be considered primarily methodological or technical on an ad hoc and mainly logical basis ("all things being equal," as we say) -- as if it has generalized meaning -- but in concrete (singular) situations, words and phrases can mean wholly different things to different people.
This matter is raised at this point since one of the most influential chapters on Q methodology, in Kerlinger's The Foundations of Behavioral Research (1986), places great importance on the proper categorization of Q statements -- as if, as in scaling, they could have only one meaning -- and also because Robert Mrtek, in his contributions to QUALRS-L on this issue, cited Kerlinger's work approvingly. Kerlinger's work is indeed important, but he attached too much weight to variance designs and their analysis, and overlooked Stephenson's admonition (in The Study of Behavior) that "it is a mistake to regard a sample as a standardized set or test of statements, any more than one can hope to regard a particular set of children as a standard sample..." (p. 77). There are many features to this subtle matter, but the bottom line is that meanings are not to be found solely in the categorical cogitations of the observer, but as well (and even more importantly) in the reflections of the individual as he or she sorts the statements in the context of a singular situation.
4. Q Sorting
In Part 3 (Q Samples), a set of N=20 statements was displayed -- abstracted from the concourse presented in Part 2 -- and it is this Q sample which is administered to participants (subjects, respondents) in the form of a Q sort. The statements are administered in the form of a pack of randomly numbered cards (one statement to a card) with which the person is instructed to operate according to some rule (called a "condition of instruction"). Typically we are interested in the person's own point of view, and so we would instruct the Q sorter to rank the statements along a continuum from "most agree" at one end to "most disagree" at the other. To assist in the Q sorting task, the person is provided with a scale and a suggested distribution. More detailed descriptions of Q sorting are to be found in Brown's Political Subjectivity (1980) and in McKeown and Thomas' Q Methodology (1988).
An example may help clarify what is involved, and for this purpose is shown the Q sort which I performed in rendering my own point of view using those statements presented in Part 3: Brown's Position -3 -2 -1 0 +1 +2 +3 16 3 1 7 6 5 2 19 13 4 8 17 9 12
15 11 10 18 14 20
Generally, the person is given the Q sample and instructed to read through them all first so as to get an impression of the range of opinion at issue and to permit the mind to settle into the situation. At the same time, the person is also instructed to begin the sorting process by initially dividing the statements into three piles: those statements experienced as agreeable in one pile, those disagreeable in a second pile, and the remainder in a third pile. The rating scale is spread across the top of a flat area (like a kitchen table), and may range from +3 to -3, or +4 to -4, or +5 to -5, depending on the number of statements. The distribution is symmetrical about the middle, but usually flatter than a normal distribution. Both the range and the distribution shape are arbitrary and have no effect on the subsequent statistical analysis, and can therefore be altered for the convenience of the Q sorter; there are, however, good reasons for encouraging the person to adhere to whatever distribution shape is adopted for the study.
The above figure shows that strongest agreement is with statements 2 and 12, which read as follows:
(2) Q methodology is a set of procedures, theory, and philosophy supporting the study of the same kind of subjectivity that is the focal point of much qualitative research.
(12) Q can give some fascinating insight into underlying philosophic structures which comprise subjective phenomena.
It is clear, therefore, that my primary concern while performing the Q sort was with the issue of subjectivity, and this is reinforced at +2 by statements 9 and 14:
(9) Cluster analysis is really something quite different and has no commitment to that subjectivity which is central to Q methodology.
(14) It allows for the interpretive study of subjective behaviors without imposing the usual biases of structured survey questionnaires.
One of the continuing frustrations that Q methodology has had to face for the more than 50 years of its existence has been the restriction of its theoretical and methodological thrust through the partial incorporation of its technical procedures -- as if all physics had to offer were its cyclotrons and behavior analysis its Skinner boxes. Hence academic psychology quite easily adopted Q sorting as a data-gathering technique, and even certain aspects of Q factor analysis, but ignored the idea of a natural science of subjectivity, and it is this protest that dominates the positive end of the above Q sort performed by one of Stephenson's students. Statement 17 punctuates the protest, like a parting remark:
(17) There is more to the method than just the technique of Q sorting. (+1)
A significant characteristic of each and every Q sort on any and all topics is its schematic nature, or what Stephenson, in his "Consciring" paper (1980), referred to as Peirce's Law (in re Charles Peirce's "Law of Mind"). There is therefore a consistency in sentiment throughout the Q sort. Under -3, for example, we see a denial of the antithesis of what is found under +3:
(16) Factor scores can be tough to come by because the correlations are of reduced rank.
(19) The frequencies in the piles must be restricted to the frequencies that would be expected if you had a normal curve, with each pile corresponding to an area of a normal curve.
Individuals unfamiliar with Q methodology are reminded that the concourse of communicability surrounding it (i.e., as expressed previously over QUALRS-L) can be highly specialized; even so, it should be easily recognized that what characterizes the positive end of the above Q sort distribution has to do with subjectivity, whereas the above statements, both scored -3, concern themselves with technicalities. This is not to say that statements 16 and 19 were found unacceptable because they dealt with technicalities: there are good technical reasons for rejecting them, but the technicalities are rooted in an appreciation of the subjectivity enbraced under +3 and +2.
Most Q technique studies involve administration of the Q sort to several respondents, but to far fewer than is the case, say, in survey research: even in studies of public opinion, samples of persons (P sets) rarely exceed 50 for reasons which will be discussed subsequently. In this particular study, we would naturally be interested in including the views of those individuals who originally contributed to the concourse -- i.e., Professors Foley, Mrtek, Kendall, et al. For the sake of time and for purposes of demonstration, I provided simulations of these individuals' views, based on their contributions to QUALRS-L. "Professor Foley's Views," for example, is as follows: Foley's Position (simulated) -3 -2 -1 0 +1 +2 +3 7 5 6 2 1 3 11 18 9 8 10 4 15 13
14 12 17 20 16 19
Without going into great detail, let us simply note that the Foley Q sort asserts that "The idea is to come up with a set of traits that characterize individuals, then compare individuals for the distribution of these sets" (no. 11, +3), and that "It is intended to get at patterning within individuals (case-wise) rather than simply across individuals (factor-wise sorting)" (no. 13, +3). Both of these were points of view which Foley espoused in his contributions to QUALRS-L (on September 23rd), and his primary concern with technical and statistical features characterizes his Q sort at the negative end as well where it is denied that cluster analysis and Q factor analysis are fundamentally different (no. 7, -3), and that "Q has never involved the correlation and factor analysis by rows of the same matrix of data that is analyzed by columns in R methodology" (no. 18, -3).
At different times over the space of three or four days, Q sorts were constructed as well to represent the views of other contributors to QUALRS-L: Professors Mrtek, Kendall, and Hofmann. For obvious reasons, a Q sort representing William Stephenson's viewpoint was also constructed for purposes of comparison with other views; and a Q sort for Fred Kerlinger, whose work on Q had been mentioned (Kerlinger, 1986); and also one representing a composite of the views of Sir Cyril Burt and R.B. Cattell, prominent exponents of factor analysis (R method) in its formative days. Also for theoretical purposes, a Q sort was constructed to represent the kind of conventional view about Q technique that one might get from a typical textbook on research methods. And finally, for reasons to which we will return, a Q sort rendition was given of what a "quantum theoretical" viewpoint about Q might be. There were therefore 10 Q sorts in all -- my own plus nine hypothetical standpoints.
As noted previously, it is unnecessary to claim that any of the above Q sorts is in any sense a "true" reflection of Foley's or Mrtek's or Burt's or anyone else's view, although I would be somewhat surprised were I to learn that I had missed the mark entirely. These Q sorts are formal models of my understanding of the points of view at issue, rendered ostensible through technique. The next installment in this series will show how these perspectives can be systematically compared. Meanwhile, in light of the discussion which has recently appeared on QUALRS-L concerning interviewing, it is important to note that a completed Q sort should be followed where possible with an interview so that the Q sorter can elaborate his or her point of view. The Q sort provides focus to the interview by indicating which of various topics in the Q sample are most worth talking about: obviously those statements scored +3 and -3 should be addressed first since they are demonstrably the most salient, but those scored 0 can be revelatory by virtue of their lack of salience.
Before concluding this section, it is well to take brief stock of what has been achieved.
(1) The Q sample is comprised solely of things which people have said, and it is therefore indigenous to their understandings and forms of life.
(2) The Q sorting operation is wholly subjective in the sense that it represents "my point of view" (whether the "me" at issue is Brown, Foley, Mrtek, or someone else): issues of validity consequently fade since there is no external criterion by which to appraise a person's own perspective.
(3) As a corollary, the factors which subsequently emerge -- factors, that is, in the factor-analytic sense -- must represent functional categories of the subjectivities at issue, i.e., categories of "operant subjectivity." All of this applies to any Q sort on any topic administered to any person in any land under any condition of instruction at any time. Subjectivity is ubiquitous, and Q methodology provides for its systematic measure.
5. Correlation
In their book on Basics of Qualitative Research (1990), Anselm Strauss and Juliet Corbin are quite explicit in distinguishing qualitative from quantitative research: "By the term qualitative research we mean any kind of research that produces findings not arrived at by means of statistical procedures or other means of quantification" (p. 17). One of the advantages of qualitative research, of course, is that it permits the systematic gathering of data which are not always amenable to quantification, but to appraise data on the basis of whether or not they have been subjected to statistical analysis is surely a case of misplaced emphasis. It is important to be able to assay the subjectivity at issue in a situation, which Q does: the fact that the resulting data are also amenable to numerical treatment opens the door to the possibility of clarity in understanding through the detection of connections which unaided perception might pass over. In Q, the role of mathematics is quite subdued and serves primarily to prepare the data to reveal their structure.
In the prior posting (Part 4), the Q sorts for my own view and that of Michael Foley (simulated) were pictured, and so it is convenient to draw on these two again to demonstrate the simplicity and subsumptive power of correlation (with apologies beforehand to those already au faix with statistics). In tabular form, the two sets of scores are as follows (where D=F-B is the difference between Foley's and Brown's scores, and D**2 is the difference squared): Item Foley Brown D=F-B D2 1 1 -1 2 4 2 0 3 -3 9 3 2 -2 4` 16 4 1 -1 2 4 5 -2 2 -4 16 6 -1 1 -2 4 7 -3 0 -3 9 8 -1 0 -1 1 9 -2 2 -4 16 10 0 0 0 0 11 3 -1 4 16 12 -1 3 -4 16 13 3 -2 5 25 14 -2 2 -4 16 15 2 -2 4 16 16 2 -3 5 25 17 0 1 -1 1 18 -3 1 -4 16 19 0 -3 3 9 20 1 0 1 1 Sum 0 0 0 220
We note, in column D, the discrepancy between the score for each item in the Foley Q sort compared to that in the Brown Q sort, and for statistical reasons that number is squared (column D2). Hence, for example, Foley gives statement no. 1 a score of +1 whereas Brown scores it -1, a difference of D=2, the square of which is of course 4. The squared differences are then summed, which, as the above table shows, produces Sum=220. Note that if the two Q sorts had been identical, each D would have been 0, each D2 would have been 0, and Sum would have been 0: when this occurs, the correlation is perfect (an extremely rare event) and is registered as r = +1.00, r being the symbol for correlation.
The specific calculation in this case is achieved first by squaring all of the scores in the Foley and Brown Q sorts and summing those squared numbers, which produces a sum of 66 for each, or 132 for the two combined. The correlation is calculated by forming the ratio of the sum of squares for Foley and Brown combined to the sum of the squared differences, and then subtracting this from 1.00. Or, in this case:
r = 1 - (Sum D**2/132) = 1 - (220/132) = -0.67
Just as a perfect positive correlation is registered as +1.00, a perfect negative correlation is -1.00, and so the correlation between Foley and Brown of r = -0.67 indicates a quite high level of disagreement, the statements which the one embraces tending to be the ones which the other rejects, and vice versa.
Foley's and Brown's are but two of the ten Q sorts at issue, and when each of the ten is compared with the others, the result is a 10x10 correlation matrix, as follows: SORT Foley 1 Mrtek 2 Kendall 3 Hofmann 4 Stephenson 5 Burt-Cattell 6 Kerlinger 7 textbook 8 quantum 9 Brown 10 1 1.00 0.17 0.79 0.76 -0.70 0.86 0.48 0.85 -0.71 -0.67 2 0.17 1.00 0.14 -0.05 0.06 0.12 0.74 0.20 -0.08 0.24 3 0.79 0.14 1.00 0.73 -0.70 0.70 0.27 0.82 -0.53 0.57 4 0.76 -0.05 0.73 1.00 -0.85 0.80 0.23 0.82 -0.77 -0.81 5 -0.70 0.06 -0.70 -0.85 1.00 -0.82 -0.17 -0.76 0.73 0.76 6 0.86 0.12 0.70 0.80 -0.82 1.00 0.39 0.82 -0.65 -0.66 7 0.48 0.74 0.27 0.23 -0.17 0.39 1.00 0.44 -0.48 -0.28 8 0.85 0.20 0.82 0.82 -0.76 0.82 0.44 1.00 -0.74 -0.67 9 -0.71 -0.08 -0.53 -0.77 0.73 -0.65 -0.48 -0.74 1.00 0.85 10 -0.67 0.24 -0.56 -0.82 0.76 -0.65 -0.27 -0.67 0.85 1.00
See Part 4 (Q Sorting) for the definitions of the 10 Q sorts.
As indicated, Brown (no. 10) correlates with Foley (no. 1) in the amount -0.67, and a quick perusal down column 10 shows that Brown correlates substantially and positively only with Q sort no. 5 (Stephenson, his mentor) and no. 9 (quantum theory); otherwise, he correlates negatively with virtually everyone else save for Mrtek, although the positive correlation in that case (r = 0.24) is insubstantial. Foley on the other hand correlates quite highly with Kendall and Hofmann.
To determine how large a correlation must be before it is considered substantial, we calculate the standard error, a rough and ready estimate of which is given by the expression 1/(SQRT(N)), where N is the number of statements (N=20 in this case) and SQRT is the square root: the value is therefore 1/(SQRT(20)) = 1/(4.47) = 0.22. As a rule of thumb, correlations are generally considered to be statistically significant if they are approximately 2 to 2.5 times the standard error -- i.e., somewhere between 2(0.22) = 0.44 and 2.5(0.22) = 0.56 (irrespective of sign). Hence in the above correlation matrix, Brown's positive correlation with Stephenson is substantial (i.e., in excess of 0.56) as is his negative correlation with Foley (i.e., in excess of -0.56), whereas his correlation with Kerlinger is insignificant (i.e., is less than 0.44).
But it is rarely the case that the correlation matrix is of much interest since attention is usually on the factors to which the correlations lead: the correlation matrix is simply a necessary way station and a condition through which the data must pass on the way to revealing their factor structure. What this involves is the subject of the next chapter.
In the meantime, it is worth stressing that the statistics associated with Q are not intended as a substitute for the obvious fact that the correlation matrix above is suffused with subjectivity, each Q sort being a transformation of a person's own vantage point, and with the coefficients merely registering the degree of similarity or dissimilarity in perspective. Moreover, although Q emerged from psychometric discussions in the 1930s, it is less and less the case that users of Q technique have need for much more than a minimal grasp of statistics. Software packages for personal computers, such as Stricklin's (1990) PCQ, or the new QMethod mainframe program nearing completion at Kent (Atkinson, 1992), convert into joy what before was drudgery, and thereby redirect attention back to the phenomenon and away from the means of its measurement.
=======================================================================[]
Q Methodology - advantages and the disadvantages of this research method Kelvin Karim
Abstract: Kelvin Karim describes how nurse researchers might use Q methodology, what it is and the advantages and disadvantages of the method.
Kelvin Karim BA, BN, RGN, Dip DN is a Community Macmillan Nurse, Walsall Community NHS Trust, Walsall, West Midlands.
There are a significant number of texts written specifically for nurses and other health care researchers. However, very few of these texts include a description of what Q methodology is and how it may be used by researchers. This article seeks to redress this gap. The advantages and disadvantages of Q methodology will also be discussed. Origins of Q methodology Q methodology was invented in 1935 by physicist and psychologist William Stephenson. Stephenson's ideas were not well received in the 1930’s when influential psychologists such as Cattell, Burt and Eysenck, were deeply rooted in experimental psychometric testing (Brown, 1998) and his ideas found more favour outside the field of psychology, especially in the USA. Brown (1998) suggests that Stephenson's thinking was ahead of his time and the proliferation of research studies involving Q methodology is testament to this. Some 50 years later, the method is increasingly used in marketing, political science, psychology, public administration and a range of more recent intellectual developments such as feminism and women’s issues.
What is it? According to Niemi (1988), Q methodology is a systematic and rigorously quantitative means for examining human subjectivity. It is essentially a method for examining intensively the subjectivity of an individual or group and in recent years has become a popular tool for researchers in social sciences. Q methodology refers to the use of Q sorting, which is a data collection technique and Q factor analysis, which is a procedure for statistical analysis. While Q sorting and Q factor analysis can be used independently, they can also be combined, enabling researchers to benefit from both qualitative and quantitative research approaches. Initially the researcher selects a set of items (stimuli) which are placed on individual Q sort cards. Collectively the cards are called a Q sort deck and may comprise of a series of statements, words, pictures, pieces of art, paintings or photographs. The Q sort deck is essentially a survey or test instrument. The researcher looking at any given subject must ensure that the full range of opinions are represented and that they are presented in an easily understandable way. Typically, the number of cards range from 60 to 100 (Polit & Hungler, 1999). Denzine (1998) cites Kerlinger (1986) as suggesting that at lease 60 cards should be used to have statistical stability and reliability. Participants are then asked to sort the cards according to certain dimensions such as approval/disapproval, most like me/least like me or lowest or highest priority. Denzine (1998) provides the example of a researcher who is seeking student perceptions on the most desirable place to study on a university campus. A Q sort deck would have to include all of the possible study locations and students would be asked to rank the cards using a desirability continuum, placed towards the middle with fewer at the extremes. Thus the middle cards are likely to contain the neutral views and extreme cards will contain the strongest views. It is the strongest views which probably contain the most important information. This is termed a forced sort. In a standard questionnaire format each question is independent of the other. However, Q methodology involves an ipsitive approach which means that each item in the Q sort deck are dependent and interrelated. The participants are less likely to respond to an item which is inconsistent with a previous item because his/her choice is likely to be restricted by the previous response. Denzine (1998) points out that in a forced sort situation, a normal statistical distribution will occur because of the ranking procedure which results in fewer cards being selected to represent the strongest held views of the participants.
Factor analysis Analysis of data obtained through Q sorts can range from the most descriptive statistical procedures such as rank ordering, averages and percentages to highly complex procedures such as factor analysis. Factor analysis, is a procedure designed to reveal the common elements in a set of items (Polit & Hungler, 1999). Put another way, factor analysis is analysis which examines inter-relationships among large numbers of variables and attempts to disentangle those relationships to identify clusters of variables that are most closely linked together (Burns & Grove, 1997). The researcher can do this manually or by using a computer programme such as Q Method software. There is a great deal of complicated mathematics which underlies factor analysis. However, it is highly fortunate that there is, as Brown (1991) suggests, little reason for the researcher to need to understand mathematics. Brown uses the analogy of the car driver not needing to know about the mechanics of a car in order to drive it. Q methodology is about appraising the subjective. Brown (1991) points out that the phenomena with which Q methodology deals is the ordinary conversation, commentary and discourse of every day life backed by a powerful statistical mechanism which can go relatively unnoticed by Q methodology users. There is no right or wrong view. Q may be used in research which seeks to explore individual or group perceptions and attitudes. This seems very appropriate in the area of health care where researchers frequently seek to elicit the view of patients and professionals. Disadvantages Conversely, it can be argued, Q methodology can be time consuming and difficult to administer. Q sorts ideally have to be done face to face and therefore obtaining a geographically diverse sample is complicated. According to Polit and Hungler (1999), the forced procedure of distributing cards according to the researcher’s specifications is the subject of criticism. Critics argue that this artificial procedure tends to exclude information concerning how people would ordinarily distribute their opinions. In addition, standard statistical procedures to Q sort data such as the interpreting data by comparing individual score with the average score for a group (normative measures) cannot be used with Q methodology due to the ipsative nature of participant responses. However, some also argue that this relatively unimportant, especially when the number of items is large (Polit & Hungler, 1999). In planning Q samples to be sorted, the researcher has to generate the statements. This can be time consuming, especially if naturalistic, that is derived from interviews or written narratives. The researcher may, however, use ready made materials. The aim of the researcher is to seek to ensure the fullest range of viewpoints in the Q-sort deck. Practically, the researcher has to make a number of decisions not only about what material to include on each card but how many piles they should be placed in and how many cards should be placed in each pile. The researcher has to decide whether the cards should be ranked in each pile. Researchers unfamiliar with Q methodology are likely to feel that the method is complicated and may lack the confidence to make these decisions. The decisions made may influence the outcomes and the researcher may find it hard to justify why each of the decisions were taken. From a participant point of view, it is not an easy exercise. The participant may struggle to understand what is required of him/her. It is also a time consuming exercise requiring the participant to think. Providing the environment and sufficient time to conduct a Q sort may be problematic for researchers and participants. The method is unlikely to be practical for use with participants who have cognitive difficulties. Researchers may also be tentative about asking people who are disabled by illness to take part. The method may not, for example, be appropriate for use with many people with a terminal illness or with learning difficulties. Popularity Q methodology offers researchers a powerful tool for systematically examining subjective data. Although there are many drawbacks in using Q methodology, there are also a number of key advantages especially for experienced researchers seeking to explore perceptions and attitudes. The method is proving popular in the social sciences and has been used by researchers in subjects as diverse as pornography, political campaigning, religion and oral history. The literature now contains around 1,500 bibliographic entries (Brown, 1986). Nevertheless, as McKeown and Thomas (1990) suggest, ‘Q retains a somewhat fugitive status within the larger scientific community’. However, at the very least, British health care researchers should have the opportunity to become familiar with Q methodology in order to decide for themselves whether or not it is appropriate to their needs. By dismissing or ignoring Q methodology, as many texts seem to, British health care researchers are being denied the opportunities that it presents.
References Brown, S.R. (1986) Q technique and method. In: Berry W.D. and Lewis-Beck M.S. (1986) (Eds) New Tools for Social Scientists. Sage, California. Brown, S.R. (1991) Q Methodology Tutorial. Kent State University, Ohio [On-line]. Available at http://facstaff.uww.edu/ cottlec/qarchive/qindex.htm Brown, S.R. (1998) The History and Principles of Q Methodology in Psychology and the Social Sciences. Kent State University, Ohio [On-line]. Available at http://facstaff .uww.edu/cottlec/qarchive/qindex.htm Burns, N., Grove, S.K. (1997). The Practice of Nursing Research: Conduct, Critique and Utilization. W.B. Saunders, London. Denzine, G.M. (1998) The Use of Q Methodology in Student Affairs Research and Practice. Student Affairs Journal-Online [Online]. Available from http://sajo.org Kerlinger, F. (1986) Foundations of Behavioural Research (3rd ed). New York. Holt, Rinehart and Winston cited Denzine, G.M. (1998) The Use of Q Methodology in Student Affairs Research and Practice. Student Affairs Journal-Online, [Online]. Available from http://sajo.org McKeown, B., Thomas, D. (1988). Q Methodology - Quantitative Applications in the Social Sciences. Sage, California. Niemi, R.G. (1988) Series Editor's Introduction. In: McKeown, B., Thomas, D. (1988) Q Methodology. Sage, California. Polit, D.F., Hungler, B.P. (1999) Nursing Research Principles and Methods. Lippincott, Philadelphia.
This page uses Creative Commons Licensed content from Wikipedia (view authors). |