BMS English Abstracts 1983-1993 (N. 1-39)
N. 1, October 1983
SOCIOLOGIE DE LA CRIMINALITE D’AFFAIRES
Karl M. van Meter
Sociology of white collar crime. The author’s intention is to present the use of two different methods of analysis showing the change in social representation they cause with data collected by the judiciary system concerning white collar crime in both Paris and the provinces. The first method, based on the use of cross-tabulations and simple statistics, characterizes the subpopulations either by direct citation or by opening of an inquiry and other differnet criteria (seriousness of the crime, damages, precedents, sentences). The second method, based on automatic classification with a null hypothesis, arrives at polythetic groups in which a majority of the individuals have a certain number of common characteristics without implying that all individuals with those characteristics are necessarily members of the group. With this method, the possibility of creating a “statistical trashbasket” characterizing a supposed type of crime as in the previous case of opening an inquiry is avoided. White collar crime, Automatic classification.
L’ELIMINATION DES MODALITES NON PERTINENTES D’ENQUETES PAR ANALYSE FACTORIELLE
Elimination of non pertinent modalities in a survey through factoral analysis. During the collection of data, the sociologist often considers the most extensive choice of modalities for the survey questions and includes many possible responses. Coding in “yes” or “no” is quite different. In order to eliminate the “noise” caused by poorly constructed modalities or codings, the author proposes, following a preliminary use of factoral analysis, to remove the “absense of observed trait” modalities, and to isolate plethoric items, weak contributions to the Khi-square and the modalities in the center of the factoral graph, thus modifying the total disjunctive table of data. With a second factoral analysis, the axes are retained, relationship between modalities remains stable, and the associations between the remaining modalities in the table improve (augmented eigen values) as well as the ease of interpreting the graph. Factoral correspondence analysis, Elimination of noise.
METHODES POST-FACTORIELLES POUR LE DEPOUILLEMENT D’ENQUETE
Post-factoral methods for survey analysis. The author’s hypothesis is that simple statistical indicators are very useful when represented in intelligent graphic form using factoral analysis graphic representation as a foundation. The surface representation of cross-tabulations (REPFAC) and the use of Khi-square percentages (PKM) require a certain expertise that varies with different populations and survey situations. The author shows how to represent the strong attractions between question modalities using as a backdrop a factoral correspondence analysis graphic. In this manner, a valued graph those peaks are the set of all question response modalities with a positive Khi-square difference. This is a post-factoral method in that it uses factor analysis to optimize a graphic representation which is often the basis of a researchers understanding of a phenomenon. Factoral correspondences analysis, Post-factoral methods.
N. 2, April 1984
LA NOTION DE REGULARITE DANS L’ANALYSE DES RESEAUX SOCIAUX
Alain Degenne, Claude Flament
The notion of regularity in social network analysis. The authors propose a language that permits one to speak simultaneiously of binary relations and relations that cannot be reduced to a set of heterogeneous couples. A binary table containing at least one 1 in each column and each row is called regular. From the structural point of view close to block modeling, the population under study is covered by different collections of relations (hypergraphs). After a remark on rules of computation for matrices representing relations, the authors develop criteria for the coherence between a network and the mapping R of one set onto another set. The sociological question remains, even in the simple case of a unique binary relation, of how to characterize groups that have antagonistic interests and do not constitue a partition. Social networks, Graphs, Regularity.
SOCIAL CLUBS, POLICY-PLANNING GROUPS AND CORPORATIONS: A NETWORK STUDY OF RULING-CLASS COHESIVENESS
G. William Domhoff
In his introduction and through an analysis of the literature, the author presents the different critiques of C. Wright Mills on the existence of a ruling political and economic elite in the USA. Although his critics do not deny the existence of such an elite, they question its hold on power by asserting that its cohesion has not been proven. The author has attempted to show that cohesion through the use of social network analysis concerning the politicians who are also powerful businessmen and frequent the same social clubs, thus confirming cohesive social behavior. The present study continues this research by analyzing a much larger data base (30 clubs, policy planning groups or foundations) and using a method of network analysis which reveals not only the distribution of ties between individuals, but also the degree of overlap between the different clubs, groups and foundations. This method also determines the centrality and the degree to overlap in relation to the general population. Both centrality and overlap indicate that the social clubs play a fundamental role in the social universe of this elite and reveal a very strong cohesiveness within that elite. Social networks, Elites, Power.
THE POWER NETWORK IN PHOENIX: AN APPLICATION OF SMALLEST SPACE ANALYSIS
Edmund M. McLaughlin
Based on a square matrix accounting for the ties between the different social institutions of Phoenix, and using a type of social network analysis, the author uses smallest space analysis to obtain the best two-dimensional representation of this matrix and thus the articulation of these institutions, each in relation to the others. The data was of two different types. First, 125 institutions including the large private companies, the banks, the insurance companies, the foundations, the large educational institutions, important clubs, and local and state government organizations. Next, the names of 1848 directors or members of these institutions were used. The first type of analysis consists of defining and measuring four different types of relationship within the population which each define a specific square matrix. The two-dimensional representation shows there exists a central node of 25 institutions including all the banks, all lawyer offices, the two large insurance companies, the municipal government and the executive of the state government. Also at the center, but outside this central node, are 32 institutions and companies. Outside this center there are 32 other institutions and businesses which form a sort of ring. The other 18 businesses without ties are not represented. The author closes by describing this structure of power in Phoenix as much more of a monolithic elite than a pluralism. Social networks, Smallest space analysis, Cliques, Elite.
INTERLOCKING DIRECTORATES IN THE TOP U.S. CORPORATIONS: A GRAPH THEORY APPROACH
John A. Sonquist, Thomas Koening
After presenting the principal theoretical orientations involved in the study of American corporate interlocks, the authors present a method for finding “cliques” within a dense network of ties. They analyze these cliques according to their centripetal force, composition, external ties, diameter, size, role of banks and insurance companies, and role of ties between the cliques. Based on this analysis and because of the high degree of inter- and intra-clique ties, the only acceptable general model seems to be that of class hegemony. An unexpected aspect was the regional character of the cliques found. Indeed, this network analysis found 32 cliques each of which has a rather clear geographical base in spite of the fact they fully participate in an extensive national, even international, network. In closing, the authors show that below this general level of analysis, the different theoretical approaches concord in producing explanations for the particular characteristics of certain cliques. Social networks, Graphs, Cliques, Elites.
N. 3, July 1983, No research articles
N. 4, October 1984
WHAT IS NOT WHAT IN STATISTICS? STATISTICAL INFERENCE REVISITED – 1984. THE ILLOGIC OF STATISTICAL INFERENCE FOR CUMULATIVE SCIENCE
Between mathematicians, statisticians and researchers in social science, there is a voluminous exchange of knowledge and conceptual tools, habitually in the direction of the former toward the latter. But this communication, which is far from perfect, has lacunas that are more or less important at each stage and particularly the utilisation of mathematical and statistical tools in the social sciences. After a preliminary in-depth analysis of this question, a text presents six classes of common statistical problems that are not yet resolved. This text is followed by some 50 brief but detailed notes on particular statistical misunderstandings and the means of avoiding them in social science research. This is followed by two recent up-datings by the author and a computer listing showing the operationalization of his remarks concerning the testing of hypotheses. Statistics, Statistical inference.
BILAN TECHNO-METHODOLOGIQUE ET PERSPECTIVES OFFERTES PAR LA NOUVELLE MICRO-INFORMATIQUE DECENTRALISEE
Technico-methodological critique and new perspectives through decentralized modern micro-informatics. Noting that increasing specialization in the division of labor in research between conceiver-researchers and server-computer technicians has lead to the overutilization of standardized techniques, the author examines the effects of the diffusion of the new generation of micro-computers in this domain. Will we witness an uncontrolled vulgarization of data analysis methods or will this new decentralized, decompartmentalized, self-directed and interactive practice encourage researchers to engage in a permanent re-examination of their data and develop a new dynamic process in the domain of methodology? Micro-informatics, Sociology of research.
STRUCTURES DES DONNEES ET STABILITE DES RESULTATS – LA TYPOLOGIE DE BASE EN SCIENCES SOCIALES
Karl M. van Meter
Structure of data and stability of results: Basic typology in social science research. The stability of results, no matter what the method, is a methodological requirement to be examined closely. For example, stability should not be affected by recoding data. One can also examine the effects of introducing noise in the data with the variables or with the subjects. Finally, one can compare results obtained by several different methods. In conclusion, one obtains stable results, “basic types”, which usually include only a small part of the total data. Recoding, Classification, Correspondence analyse, Principal component analysis, Dynamic clouds, Stability.
N. 5, January 1985
INDIVIDUS FICTIFS EN SUPPLEMENTAIRES DANS UN PLAN FACTORIEL
Supplementary fictious individuals on a factoral analysis graphic. In factoral correspondence analysis, the supplementary individuals technique is well known. Here the author proposes an application of the technique which in a simple manner creats fictious individuals. This technique is then used to facilitate the understanding and interpretation of analysis results. Correspondence analysis, Supplementary individus.
N. 6, April 1985
BAROUF A BOMBACH
Trouble at Bombach. This text presents the paradox of a two-dimensional cross-tabulation in which similar effects are witnessed in two subpopulations but opposite effects are witnessed at the level of the general population. Cross-tabulation tables, Simpson’s paradox.
BAROUF A BOMBACH (SUITE)
Trouble at Bomback (continued). This article concerns the “Trouble at Bombach” paradox and shows that the paradoxical situation is only possible when the largest proportion of a subpopulation is superior to the smallest of another subpopulation. Cross-tabulation tables, Simpson’s Paradox.
BAROUF A BOMBACH (SUITE MONTPELLIERAINE)
Trouble at Bombach (Montpellian follow-up). This article describes the presentation of the Bombach paradoxe as a exercise in a Montpellier university course in sociology. Cross-tabulation tables, Simpson’s paradox, Teaching statistics.
BAROUF A BOMBACH: SUITE PROBABILISTE
Trouble at Bombach: probabilistic follow-up. This article concerns the probability of encountering a Bombach paradox situation. Cross-tabulation tables, Simpson’s paradox.
OUBLI ET PARADOXE
Forgetfulness and paradoxes. Paradoxical cases (such as “ecological paradoxes” or “Bombach paradoxes”) are found in cross-tabulation tables with contradictory information between the margins and the columns. Multivariate analysis often forgets that the weighting of margins should be taken into consideration and that these paradoxes are often only the manifestion of this forgetfulness. Cross-tabulation tables, Ecological paradox, Multivariate analysis.
L’AXE DES PARADOXES DANS LA METHODOLOGIE SOCIOLOGIQUE
Karl M. van Meter
The axis of paradoxes in sociological methodology. In sociological research, researchers choose with care their variables, their subjects, the coding used and the methods of statistical analysis to treat the collected data. What remains imprecise are the preceeding stages (the transformation of information into survey data) and following stages (the interpretation of survey results). These are aspects that methodological research must examine further. Methodology.
N. 7, July 1985
BRIEF OVERVIEW OF SOVIET LITERATURE ON MATHEMATICAL METHODS IN SOCIOLOGY (1973-1983)
V. G. Andreyenkov, Ju. N. Tolstova
Presentation of a bibliography of some 250 articles and monographs concerning the application of mathematical techinques used in sociological data analysis. The principal aspects treated are the following: analysis of qualitative or quantitavie data; preference analysis; classification analysis; proximity and spatial analysis of variables; causal analysis; problems of measure. USSR, Mathematical methodas.
N. 8, October 1985
APPORT DE LA METHODE A.S.I. DANS UNE ANALYSE SECONDAIRE
Contrubution of the method A.S.I. in a secondary analysis. Presentaton of an originale descriptive method: the Structural Analysis of Interference” (A.S.I.). Using the algebric struture of a distributive lattice of n generators, this method calculates a system of coefficients intended for the local analysis of complex contingence table (particularly of more than two dimensions). The “joint rarity” coefficient is the most useful. It measures the distance, square by square, of the number of sujects observed and the number anticipated. The values of this coefficient are directly interpretable in terms relationships, within the limitations imposed by the size of the mathematical calculations. The author also proposes another system of “weighted interference coefficents” which offer an appreciation of the fiability of the observed results in the sample. A detailed example of the use of the method is also presented. Descriptive methods, Data analysis, Contingency tables, Interference, Interaction, Rarity coefficients, Boolian algeber, Ddistributive Lattices, Entropy.
N. 9, January 1986
LES ASSOCIATIONS SOCIO-COGNITIVES MOBILISEES PAR UN QUESTIONNAIRE D’ASSOCIATION DE MOTS
Social-cognitive associations implicated in a questionnaire on word association. An article of epistemological reflexion on the statistical results of a word association questionnaire concerning economic terms and used in an employee survey. The article concerns the formalized hypotheses of the questionnaire, the meaning of these questions, the typologies of association produced, and the theory of representations that implicitly understood to produce cognitive results and not to reveal a pre-established representation. Smallest space analysis, Graphs, Ties, Economic representations, Configuration of cognitive space, Linguistic semiology.
DIMENSIONS OF ASSOCIATION IN SOCIOLOGY : AN ORGANIZATIONAL MAP OF AN ACADEMIC DISCIPLINE
Charles L. Cappel, Thomas M. Guterbock
A study of the organization of American sociology through the volontery membership of sociologiest between 1980 and 1984 ub the American Sociological Association. The authors describe by multidimensional scaling and hierarchical classification methods the importance and proximity between disciplinary specialties using data concerning joint membership in any of the 32 different specialized section of the Association. Sociology of science, Epistemology, American Sociological Association, Multidimensionnal scaling, Hierarchical classification.
ENUMERATING AN INNER CITY POPULATION: A RESEARCH REPORT
R. Bruce Wiegand
To estimate the numbers of homeless, which is a very changing and unstable population, the author proposes several extensive single time “censuses”. He presents the example of direct observations repeated once during each of the fous seasons in Nashville. He criticizes the usual techinques of estimation based essentially of appreciations made by informed experts of the Census, Homeless.
N. 10, April 1986
A PROPOS DE METHODES: EFFETS D’OPTIQUE, HEURISTIQUE ET OBJECTIVATION
Concerning methods: optical effects, heuristics and objectivization. Going beyond the division between qualitative and/or quantitative methods, the author, using the example of a particular survey, presents a “middle of the road” methodology. He recalls that most surveys are minimally present at the pole to which they do not contribut (individual biographies used to prepare a standardized questionnaire or the multiplication of questions in a questionnaire in order to be more comprehensive). Thus, the classic duality between comprehensive and generalizing approaches is not tied to any particular method which is qualitative or quantative. Choices between the two approaches can occure at any time during a study. Approache by abstraction, Comprehensive approache , Qualitative methods, Quantitative methods, Biographies, Pre-questionnaire, Statistical Surveys, Multivariate aalyse, Factoral correspondence analyse.
QUALITATIVE “VERSUS” QUANTITATIVE METHODS IN SOCIAL RESEARCH
Thomas P. Wilson
The perenial controversy over quantitative and qualitative methods is addressed from an empirical point of view rather than as a purely methodological or epistemological question. Consideration of empirical findings leads to rejection of the prevailing views that social science is either nomothetic or idiographic and that a significant methodological distinction can be drawn between quantitative and qualitative methods in social research. As a consequence, quantitative and qualitative methods are seen as inherently complementary, and the question of objectivity in social science is viewed, not as a matter of following “correct” procedure, but rather as an issue dealt with as a pratical matter by researchers in terms of internal and external coherence. Qualitative methods, Quantitative methods, Logical positivist, Social categories, Social interaction, Context-free, Context sensitive mechanism, Reflexive context.
NOTE SUR LES TENDANCES METHODOLOGIQUES DES SCIENCES SOCIALES AU BRESIL
Note on the methodological tendancies in Brazilian social sciences. Work associated with major theories is much more valued than work on the tools of research, thus methodology is often relegated to a subalternate position. The major tendencies are the use of empirical and quantative techniques and the use of participative qualitative approaches. Statistical techniques are not oftenly used although micro-computers have tended to demystify “quantitatism”. Qualitative methods, Quantitative methods, Social science, Brazil.
N. 11, July 1986
ON THE LEVEL: MEASUREMENT SCALES AND SOCIOLOGICAL THEORY
This paper examines the variety of interpretations of how conceptualization is taken to shape sociological measurement. Part I offers a critique of the way in which measurement properties are currently established and shows that theory and measurement are still understood as broadly separated enterprises. Part II attempts to develop an interpretation of measurement consonant with post-empiricist philosophy and advocates the reworking of sociological theory so that it can play a more prominent role in grounding measurement systems. In the light of this revised understanding of enquiry, Part III is a post-script and post-motem on “level of measurement”. Measurement scales, Post-empiricist philosophy, Sociological concepts, Social class measurement, Social mobility theory.
DIALECTICS, MARXISM AND SOCIAL SCIENCE RESEARCH
Pauline Marie Vaillancourt
First, this paper examines different conceptions of the dialectic within the Marxist tradition and shows that one, in particular, is more consistent with empirical research than the others. Second, the view of certains Marxists, that empirical research techniques are incompatible with a dialectic approach, are also discussed and rejected. Third, specific research tools common to modern social science are examined to detect if they incorporate assumptions that contradict dialectical explanation. Last, in conclusion, rather than repudiating dialectics, social scientists should reevaluate its potential for studying dynamic phenomena. Dialectic, Marxism, Empirical research, Dynamic phenomena, Time series, Panel analysis, Longitudinal data, Symetrical causality, Feedback, Interaction, Recursive structures, Hierarchical model, Path analysis, Stochastic process.
N. 12, October 1986
LES MODELES LOG-LINEAIRES POUR L’ANALYSE DES TABLEAUX DE CONTINGENCE
Log-linear models for the analysis of contingency tables. Based on an example of the consumption of psychotropic medecins, the author explains how to select the models of interactions between variables, measure the ajustment of the model and interpret the parameters in order to estimate the relative important of the effects of the different variables and their interactions. Log-linear models, Cross-tabulations tables.
L’OBTENTION DE CARTES STRATEGIQUES DE LA RECHERCHE SCIENTIFIQUE A PARTIR DE L’ANALYSE DES MOTS ASSOCIES DANS LES FICHIERS DOCUMENTAIRES
Jean-Pierre Courtial, Jean Juan
Obtaiing strategic maps of scientific research based on the associated word analysis of documentary data files. Using a documentary data file on artificial intelligence, the authors try to graphically represent the proximity between concepts (and then between concepts and countries). The first method is based on the quantification of the association between two keyword using the number of their observed cooccurence in the documentary file. One then aggreagates the stronger pairs and the words associated with them in order to obtain groups of of a determined number of words with ties that are more or less strong. The representation of this graph is optimized (leximappe graph). A second method consists of to each frequent term a boolean vector of presence/absence of cooccurence of less frequent terms in the lexique. The representation of the results can be done by the previous method or by a factoral correspondence analysis. Documentary analysis, Leximappe graphic, Sociology of science.
UNE NOTE SUR LES “TYPES DE BASE” EN ANALYSE HIERARCHIQUE
Yannick Lemel, Daniel Verger
A note on “basic types” in hierarchical classification anlaysis. The authors compare results from a dozen hierarchial classification analyses of the same objects. They find that the groups of objects are practically the same in every case but the larger groups are not stable from one technique to another. If “basic types” exist, they are of very small size. Basic types, Hierarchical classification anlysis.
N. 13, janvier 1987
ON THE METHODOLOGY OF DISCOURSE ANALYSIS AND CONTENT ANALYSIS1
Throughá examples of analysis of economic texts by aá formal ì methodá ofá sociolinguistics,á the author revealsá notá onlyá the ì relations between economists and state power, but also shows that ì theseá examples themselves are part of theá relations.á 1Discourse ì analysis,áá contentá analysis,áá methodology,áá sociolinguistics, ì economy, education, state power.1
METHODOLOGICAL DISCOURS I: SOCIOLOGICAL RESEARCH AND SOCIAL REPRESENTATION OF DEVIANCE IN THE STUDY OF IDEOLOGY
Karl M. van Meter
Recent research in social representation and sociological methodology, and in particular the analysis of French governent sponsored sociological research projects (AlPs) and French scientific literature in sociology, show that social representation and discourse on methodology play an important role in determining the outcome of sociological research. Social representation, methodology, discourse analysis, French sociology, deviance, ideology.
CLASSIFICATION DESCENDANTE HIERARCHIQUE ET ANALYSE LEXICALE PAR CONTEXT – APPLICATION AU CORPUS DE POESIE D’RIMBAUD
Hierarchically descending classification and lexical analysis by context: application to the corpus of A. Ri baud’s poetry. Using a lexical analysis by context and a hierarchically descending classification method, the author examines the corpus of Rimbaud’s poetry. Two different codings of the corpus with the use cf the same method of analysis, plus other methods also, furnish a means of judging the stability of the results obtained. Lexical analysis, Arthur Rimbaud, methodology, hierarchically descending classification analysis, coding.
Les Structures de l’opinion en 1985 – Enquêtes et méthodologie
The structure of public opinion in France in 1985 – Surveys and methodology. By the use of a systematic selection process for questionnaire construction and a rigorous methodology controlling for the known sources of survey variability and bias, these successive surveys since 1977 furnish a detailed representation of how French public opinion is stably structured around a specific field of conflicts. Principal component analysis is the method of data anaylsis used here. Public opinion, surveys, conflicts, principal components analysis, bias, methodology.
N. 15, juillet 1987
IDEAL TYPES, SYNDROMES, AND POLYTHETIC CLASSES: THE OPERATIONALIZATION OF CROSS-CLASSIFICATION ANALYSIS IN BEHAVIORAL SCIENCE RESEARCH
Karl M. van Meter, Marten W. de Vries, Charles D. Kaplan, Chantal I. M. Dijkman
By a detailed analysis of the concept of classification and nomenclature of some behavior problems, the authors refine the concepts syndrome, class, model and ideal type, specifying the close relations between them. The method of cross-classification provides a way to operationalize these considerations. The experience sampling method (ES), which introduces variables for mental status and the immediate environment, provides a new research tool. Together, these developments form a research methodology currently used in two major research programs on behavior, one on anxiety, the other on the use of drugs. The combination of these two techniques reveals new aspects of the relationship between syndromes, analysis and models. Syndromes, Polythetic Classes, Models, Cross-classification, Ideal types, Operationalization, Anxiety, Drugs, Experience sampling (ES).
N. 16, octobre 1987
GEOMETRIE ET LANGAGE – LA STRUCTURE DES MODELES EN SCIENCES SOCIALES ET EN SCIENCES PHYSIQUES
Geometry and language: the structure of models in social sciences and in physical sciences. The first part argues that in physical sciences models are essentially based on visual or reified geometry while in the social sciences, models are often based on language constructions and, in particular, metaphors. The second part examines in detail the relationship between research in the physical sciences and in the social sciences, specifically according to their ontological reference to the models they use. Everyday language is the physical sciences is analyzed as well as three theories of metaphor. The author develops the idea that in the social sciences, there is a deterministic relationship between metaphoric models and social ontology. Models, physical sciences, social sciences, geometrical representations, language metaphors.
N. 17, janvier 1987
THE METHODOLOGY OF THE ATLAS OF CORPORATE INTERLOCKS
Joel H. Levine
The interlocks between boards of directors of some 400 of the world’s largest corporations are studied with methods of network analysis and, in particular, with the original methods of centroid scaling and frequency reconstruction scaling. The former furnishes stable global maps of these interlocks and discerns five groups of corporations: the French group; the Swiss group; the German group; the Dutch group; the British-Canadian-American-South-African group. The latter method furnishes very detailed maps of the internal structure of these groups and the ties between these groups. This article consists of the methodological sections of the author’s book, Atlas of Corporate Interlocks, Worldnet (Box A-201, Hanover NH 0375), 1985, 2 volumes, 501 p, 495 US$. Social network analysis, interlocking directorates, corporations, centroid scaling, frequency reconstruction scaling.
LE SEMINAIRE “UN NIVEAU INTERMEDIAIRE – LES RESEAUX SOCIAUX”
AFARS (Association Française des Analystes de Réseaux Sociaux), BMS
The Seminar “An Intermediate Level – Social Networks”. This article includes a presentation of this international seminar, a description of the published proceedings and a detailed abstract of each of the eleven presentations given during the seminar. Social network analysis, France.
EVALUATION OF SOCIOLOGY SERIALS BY A SELECT SAMPLE OF SOCIOLOGISTS
Leo P. Chall (Sociological Abstracts)
In this original survey, 123 major English-speaking sociologists, including many editors of sociology serials, ranked by importance of scientific contribution, 90 sociology serials published in English. An analysis of these data shows the structuring of this domain and an important difference between annuals and other serials. Also, the non-American English-speaking subpopulation behaves in a manner different from the American subpopulation. Sociology of sociology, scientific publishing, American sociology, English language sociology, scientific prestige.
N. 18, avril 1988
TYPES OF INADEQUATE INTERVIEWER BEHAVIOUR IN SURVEY INTERVIEWS;
THEIR CAUSES AND EFFECTS1
Johannnes van der Zouwen, Wil Dijkstra
The authors present a model of the interaction between interviewer and interviewee during the survey interview process. On this theoretical basis, survey interviews were recorded and analysed, revealing four types of inadequate behaviour: the adaptation of questions for the interviewee by the interviewer; the interviewer’s inattentiveness; the interviewer’s choosing of a response “on the behalf” of the interviewee; the hinting by the interviewer to clarify the interviewee’s response. These different types are thoroughly analysed, showing that it is especially hinting that distorts survey interview data. Interviewer, interviewee, interview, survey, questionnaire, behaviour.
LES ENQUETES D’OPINION ET LA RECHERCHE EN SCIENCES SOCIALES: RESUMES DES INTERVENTIONS LORS DE LA JOURNEE D’ETUDE ORGANISEE PAR LA SOCIETE DES AMIS DU CENTRE D’ETUDES SOCIOLOGIQUES EN HOMMAGE A JEAN STOETZEL, LE 29 FEVRIER 1988
Opinion polls and social science research. Summaries of presentations during the day of study in memory of Jean Stoetzel, organised by the Society of friends of the Centre of sociological studies on 29 February 1988. This day of study consisted of six presentations: Hélène Riffault, “Public perception of opinion polls”, which presents and analyses a detailed survey on the public perception of opinion polls; Jacques René Rabier, “The comparative and diachronic analysis of survey data”, which presents the results of European Communities Commission surveys concerning well-being, confidence between people, and national cultures; Jean Paul Grémy, “The problems of secondary analysis”; Guy Michelat, “Opinion polls in the study of political behaviour”; Yannick Lemel, “Opinion polls in the study of socio-economic behaviour”; François Chazel, “Opinion polls in the study of cultural practices”. Jean Stoetzel, IFOP, opinion polls, surveys, secondary analysis.
MATHEMATIQUE ET SCIENCES SOCIALES: HEURES ET MALHEURS 1960-1980 – COMPTE RENDU DE LA JOURNEE D’ETUDE DE LA SOCIETE FRANCAIS POUR L’HISTOIRE DES SCIENCES DE L’HOMME (SFHSH), LE 16 JANVIER 1988
Mathematics and social sciences: Highs and lows, 1960- 1980. Summary of the day of study organized by the French Society for the history of the human sciences on 16 January 1988. This day of study included an initial presentation by Jacqueline Feldman, and presentations by Bernard Monjardet on mathematics in the social science curriculum, by Marie-Ange Schiltz on mathematics in sociology, by Marion Selz-Laurière on mathematics in ethnology, by Benjamin Matalon on mathematics in psychology, by Denise Pumain on mathematics in geography, by Sylvie Rimbert on mathematics in cartography, and by Jean Philippe Genet on mathematics in history. Mathematics, social sciences, teaching mathematics, sociology, ethnology, psychology, geography, cartography, history.
N. 19, July 1988
AN INTRODUCTION TO THE GRADE OF MEMBERSHIP CLASSIFICATION MODEL
David J. Jackson, Max A. Woodbury, Kenneth G. Manton
The Discrete GOM Model at a conceptual level is a model with enormous theoretical appeal as a psychometric model. While it was initially developed as a solution to problems posed in medical classification, social scientists should be quick to realize its many potential applications. This is especially true for sociologists who have been captivated by the intuitive appeal of typological theories, but have been limited by inadequate methodological tools. The GOM models have a number of interesting statistical properties and have proved to be an effective theoretical and data analysis tool in a number of disciplines. The true value of the GOM model will be known by the capacity to make theoretical and practical sense out of data. This paper provides an introduction to GOM Analysis in terms of a heuristic analogy and a concrete sociological example of a two pure type model of parental values. The primary emphasis is on providing a conceptual perspective, rather than a detailed mathematical and statistical specifications. Classification, Grade of Membership, Fuzzy Sets, Maximum Likelihood, True Score Models, Latent Structure, Latent Class Analysis.
N. 20, octobre 1988
ATTITUDES DES SOCIOLOGUES A L’EGARD DE LA NOTION DE GENERALISATION DES RESULTATS: QUESTIONNAIRE SUR L’INFERENCE STATISTIQUE
Sociologists’ attitudes toward the generalization of results: a questionnaire on statistical inference. A questionnaire was presented to 19 sociologists. It required the use of statistical inference based on results obtained from numerical data from a hypothetical study. The questionnaire responses were compared with a Bayesian method in order to study the sociologists’ attitudes in making a statisitical inference. The responses were very heterogeneous and many subjects were quite reserved about working with results based only on numerical data with no supplementary information. Statistical inference, Bayesian methods, numerical data, generalisability, social science research.
INTRODUCTION A L’UTILISATION DU LOGICIEL D’INFERENCE BAYESIENNE SUR LES FREQUENCES (IBF2XK)
Jean-Marc Bernard, Philippe Cibois
Introduction to the use of the computer program “Bayesian Inference based on Frequencies” (IBF2XK). With several different examples of frequency distributions, the authors show how to use this microcomputer program of bayesian analysis to answer typical questions of statistical inference that arise in social science research. Examples include inferences from one or two frequency distributions, significance tests, variance from independance, and confidence intervals. Bayesian analysis, statistical inference, frequency distributions, significance tests, variance from independance, confidence intervals.
N. 21, octobre 1988
INTEGRATION AND DIFFERENTIATION IN INTERNATIONAL SOCIOLOGY: CLUSTERING OF ISA RESEARCH COMMITTEES
Slavko Splichal, Anuska Ferligoj, Zdravko Mlinar
The growing differentiation within the International Sociological Association (ISA) requires different forms of integration among Research Committees (RCs). The aim of this paper is to reveal some patterns of integrational versus differentiational activities and produce an organizational and conceptual typology of RCs. The problem is approached by different clustering methods: Ward’s hierarchical clustering, local optimization methods and hierarchical clustering with relational constraints for different types of connectivity between units in clusters. International Sociological Association, Sociological Specialization, Classification Analysis, Typological Analysis, Classification with Constraints.
COMPLEMENTARITE METHODOLOGIQUE DES APPROCHES QUANTITATIVE ET QUALITATIVE DE LA RECHERCHE DANS LE CHAMP DE LA TOXICOMANIE
France Rudolph Ingold, Sylvie Ingold
Methodological complementarity of quantitative and qualitative research approaches in the study of toxicology. With the intention of showing the complementarity of quantitative and qualitative methods, this article présents three studies done in Paris and the Paris region between 1980 and 1986 in the domaine of toxicology. The first study analyzes 793 questionnaires of first requests of assistance made to specialized institutions, complemented by an ethnographic analysis. The second study analyzes 224 questionnaires and many non-directive interviews with imprisoned drug users. The last study concerns statistics of deaths attributed to the use of drugs. Quantitative Methodes, Qualitative Méthodes, Methodological Complementarity, Toxicology.
N. 22, March 1989
ANALYSIS AND QUESTIONNAIRE OF THE SURVEY “PUBLIC OPINION IN THE SOVIET UNION AND WEST GERMANY”
Following the previous joint Soviet-American and Soviet-French public opinion polls, the author and the Institute of Sociology of the Academy of Sciences of the USSR organized the first Soviet-West German public opinion poll during the October 1988 visit of Helmut Kohl to the Soviet Union. The questionnaire is presented along with a brief analysis of the results. USSR, West Germany, Public opinion polls.
ENGLISH AND FRENCH ABSTRACTS OF RESEARCH ARTICLES PUBLISHED IN THE BMS FROM ISSUES 1 TO 12: PART I
N. 23, June 1989
L’ANALYSE DES DONNEES MULTIDIMENSIONNELLES PAR LE LANGAGE D’INTERROGATION DES DONNEES (LID): AU DELA DE L’ANALYSE DES CORRESPONDANCES
Jean-Marc Bernard, Brigitte Le Roux, Henry Rouanet, Marie-Ange Schiltz
Multidimensional data analysis with LID data interrogation language. Chapter I describes the three stages which lead to the analysis of a questionnaire by the language LID: correspondence analysis, data structuring, data exploration with LID. Chapter II describes the language LID and the interface program INTERF between EyeLID-1 and ADDAD. Chapter III presents two concrete examples of application: the first is the reanalysis of a British survey; the second is the analysis of a French survey. Chapter IV, in the framework of a ternary frequency table, presents a study of data interaction. Multivariate Data Analysis, Factoral Correspondence Analysis, Data Interrogation, LID.
ENGLISH AND FRENCH ABSTRACTS OF RESEARCH ARTICLES PUBLISHED IN THE BMS FROM ISSUES 1 TO 12: PART 2
N. 24, September 1989
CROSS-CULTURAL AND INTERNATIONAL COMPARATIVE RESEARCH OF SOCIAL MOBILITY
Rudolf Andorka (Karl Marx University of Economics, Budapest)
Initially presented at the Research Council meeting of the International Sociological Association in October 1988 in Ljubljana, Yugoslavia, this article traces in detail the development and evolution of social mobility studies, and their comparison between countries, from their conception in the 1920s and through successive “generations” with their concomitant methods which have included inflow-outflow rates, path analysis, log-linear analysis, and life history methods. Social Mobility, Comparative Research, Methodology, History of Methods.
DECORRELATION DES VARIABLES CATEGORIELLES
Jacques A. Zighera
Decorrelation of categorical variables. Using data from a survey of family structure and female occupation, the author shows how categorical variables can involve co-linearity as with continuous variables. Through the use of log-linear analysis and analysis of variance, these problems and traps of co-linearity can be resolved and avoided. Co-linearity, Categorical Variables, Log-linear Analysis, Analysis of Variance.
N. 25, December 1989
CONSTRUCTING TYPOLOGIES THROUGH CLUSTER ANALYSIS
Kenneth D. Bailey
The lack of a statistical model has been said to make clustering methods inadequate for finding underlying types. It is shown that clustering, while lacking a statistical model, does have a well developed logical model. Monothetic, polythetic, and fully polythetic types are defined. These are logical definitions, based on the logic of classes. The goal of typology construction is stated. The relationship between the logical model and a statistical model based on the mixture problem is discussed. It is shown that typology construction (via clustering or any other method) is unnecessary for studying a small number of variables, but is essential for a large number of variables, and that clustering is the only method that can adequately construct the necessary typology. Cluster Analysis, Monothetic Classes, Polythetic Classes, Typologies, Statistical Models, Logical Models.
JACKKNIFE AND BOOTSTRAP ESTIMATES FOR CLUSTER VALIDATION: EXAMPLES USING U.S. OCCUPATIONS
The author discusses the capability and implementation of jackknife and bootstrap simulation techniques for estimating bias in statistical procedures, particularly in placing confidence intervals on the quality of cluster analysis results. The author examines the reproducibility of a classification scheme by applying associated statistical procedures to access the stability of the clustering solution. Using the example of U.S. occupation data, the author compares bootstrap simulation to discriminant analysis results concerning cluster validity, and provides a listing of SAS commands to construct bootstrap samples. Cluster Analysis, Simulation, Jackknife, Bootstrap, Discriminant Analysis, Validity, Confidence Intervals, U.S. Occupation Data.
N. 26, March 1989
ECLAIRER LE VOCABULAIRE DES QUESTIONS OUVERTES PAR LES QUESTIONS FERMEES: LE TABLEAU LEXICAL DES QUESTIONS
Clarifying the Vocabulary of Open Questions by Using Closed Questions: The Lexical Table of Questions. The “Lexical Table of Questions” which cross-tabulates the vocabulary of an open question with the closed questions of a survey is a tool of data analysis which permits the clarification of each type of question by the other. An application using factorial correspondence analysis of such a table is presented concerning an opinion survey on the reform of French spelling. Open questions. Lexical Table of Questions. Factorial Correspondence Analysis. Content Analysis.
ALCESTE: UNE METHODOLOGIE D’ANALYSE DES DONNEES TEXTUELLES ET UNE APPLICAITON – “AURELIA” DE GERARD DE NERVAL
ALCESTE – A Methodology of Textual Data Analysis and an Application: Aurélia by Gérard de Nerval. Beginning with a cross-tabulation with different all sentence fragments in rows and a selected vocabulary in columns for a specific corpus, the author presents: the methodology, including principle concepts and objectives of this form of analysis; the technique, the ALCESTE computer program of automatic classification based on resemblance or dissimilarity; and an application, the analysis of Gérard de Nerval’s text Aurélia. The analysis distinguishes three types of fragments which are described and analyzed further. ALCESTE. Textual Analysis. Gérard de Nerval. Aurélia. Hierarchical Descending Classification.
N. 27, June 1990
MAUX DE L’ECONOMIE, MOTS DES ECONOMISTES
Economic Troubles, Economists’ Words. Economic methodology has greatly separated discourse (hypotheses, theory, body of thought) and action (essentially represented by applications or descriptive monographs). In this study, with the greatest care associated with original attempts, we try to give a synthetic representation over a long period of time of the evolution of economic literature by using classic methods of data analysis. No matter what type of results are obtained, this approach remains based on the hypothesis that each literary production can be taken as a simple statistical object. The scope of this study covers work published in French on work economy from 1960 to 1980 and reveals three tendencies: (1) an decrease in accordance with the real subjects studied (job segmentation and increasing unemployment); (2) a profound desire to better assure the link between analysis of economic policy and a body of thought; and (3) a certain divergence according to dissatisfaction or failures encountered in studies by authors of the two preceding categories. Sociology of Labor. Economic Policy. Text Analysis. Social Representation of Economy.
METHODOLOGICAL ISSUES IN CONVERSATION ANALYSIS
Paul ten Have
Conversation Analysis (CA), a research tradition that grew out of ethnomethodology, has some unique methodological features. It studies the social organization of ‘conversation’, or ‘talk-in-interaction’, by a detailed inspection of tape recordings and transcriptions made from such recordings. In this paper, the author describes some of those features in the interest of exploring their grounds. In doing so, he discusses some of the problems and dilemma’s conversation analysts deal with in their daily practice, using both the literature and his own experiences as resources. He presents CA’s research strategy as a solution to ethnomethodology’s problem of the ‘invisibility’ of common sense and describe it in an idealized form as a seven step procedure. The author also discusses some of the major criticisms leveled against it and touches on some current developments. Conversation Analysis is a disciplined way of studying the local organization of interactional episodes, its unique methodological practice has enabled its practitioners to produce a mass of insights into the detailed procedural foundations of everyday life. It has developed some very practical solutions to some rather thorny methodological problems. As such it is methodologically ‘impure’, but it works. Conversation Analysis. Ethnomethodology. Interaction Episodes. Quantification. Coding.
SYSTEMATIC CODING IN QUALITATIVE RESEARCH
Anselm L. Strauss
In the framework of Grounded Theory and qualitative research, the author uses a study of medical technology and medical work to examine the supposedly “self-evident” procedure of coding field observations to show how in concrete terms different categories of coding and variables can be elaborated. He treats open coding, axial coding, selective coding, and coding for structural and interactional relationships. He ends with several rules of thumb concerning coding procedures. Grounded Theory. Coding. Field Observation. Qualitative Methodology.
N. 28, September 1990
IMAGES DE LA RECHERCHE SOCIOLOGIQUE ACTUELLE
Ghislaine Chartron, Alain Degenne, Marie-Odile Lebeaux, Lise Mounier
Images of Current Sociological Research. Three sociological associations participated in 1988 in the production of a directory of French-speaking sociologists: Sociologie française et francophone. This directory includes research topics – presented as text in an open question – and key words. After an initial compilation of the vocabulary (with Lexinet), 1460 entries were the object of a statistical classification analysis (with Leximappe) in order to reveal the research problematics covered by the corpus. A statistical indicator of coherence and a more subjective evaluation of thematic coherence showed: research specialties which were both narrow and very homogeneous, more open specialties organized around a thematic core, and finally traditional specialties which persist. It is more the objects of study than the concepts employed which structure the domain. It is a picture of a rich and active sociology closely tied to social problems. Text Analysis, Lexinet, Leximappe, Structure of French Sociology, Classification.
RELIABILITY AND VALIDITY APPRAISAL THROUGH STRUCTURAL EQUATION MODELS
From an historical overview of reliability and validity appraisal it is concluded that too narrow a view of validity has been taken by researchers working in the tradition of structural equation modelling, especially as applied to the classical factor model and its multi-trait multi-method extension. The partitioning of error variance in the approach between substantive effects, and method effects and its relation to unreliability and invalidity is explained; the alternative simplex interpretation of the relations among a set of indicators is then presented. An empirical example is used to demonstrate the dependence of reliability and validity estimates on the theoretical context in which they are made. It is concluded that, following Cronbach, the meaning of the term validity should be restored to encompass all the evidence that the theoretical properties assigned to a measurement are achieved and that the specific biases brought about by method effects and other confounding variables should be referred to in terms of measurement accuracy. Reliability, Validity, Structural Equation Models, Measurement.
RELIABILITY: A DISSENTING VIEW
“Validity” is one of the few metaterms on whose meaning a certain degree of consensus has been reached by the social science community which currently tends to define it as a (more or less informed, more or less consented) subjective judgment on the degree of semantic correspondence between a concept A and another concept B, taken as an indicator of A. As such, validity cannot be measured, and alleged coefficients of validity are actually measuring something else; i.e., some form of statistical r between vectors. Two consequences ensue: (1) reliability cannot be measured, since there is no way of directly “apprehending” an individual’s state on a property; (2) reliability cannot be a property of a variable, or indicator, or operational definition, or of anything pertaining to a plurality of individuals. Validity, Semantic Correspondence, Reliability, Variables.
N. 29, December 1990
THE RELATIONSHIP BETWEEN MODE OF ADMINISTRATION AND QUALITY OF DATA IN SURVEY RESEARCH
Johannes van der Zouwen, Edith D. de Leeuw
In a series of meta-analyses on the outcomes of “mode experiments”, we found small, but statistically significant, and consistent, effects of the method of data collection (“mode effects”) on the quality (i.e., the representativity, completeness, and accuracy) of the data gathered in surveys. On several indicators of data quality, the face-to-face interview scores slightly better than the telephone interview, but these differences are gradually disappearing over time. Mail surveys usually have lower response rates and less item response than face-to-face interviews and telephone interviews. Mail surveys result in more accurate, less biased, answers than both forms of interviewing, especially when sensitive or embarrassing questions are being asked. In order to give an explanation for these mode effects, we distinguished three “mode characteristics”, or intervening variables (the “persuasion power” of a mode, the complexity of the task for the respondent, and the degree of control over the question-answer process, i.e., control over role-related behavior of respondents and interviewers), and one conditional variable (the degree to which the response categories of the questions differ with respect to their social desirability). These variables were “inserted” in a model which has as independent variables the three modes concerned, and as dependent variables the response rate, and the completeness and accuracy of the responses. Face-to-face interviews, Telephone interviews, Mail surveys, Data quality, Mode effects, Meta-analysis.
WEIGHTING PROCEDURES FOR ETHNOGRAPHIC RANDOM SAMPLES
In random sampling, the researcher may not exercise any personal preference in the selection of respondents. In contrast, an ethnographer chooses respondents for their informative value. This paper proposes “ethnographic random sampling,” in which a sample is selected according to ethnographic methods, but with random selection from strata composed of similar potential respondents. This synthesis retains the adaptive initiative of ethnography while providing the generalizability of random sampling. The sampling frame consists of all potential respondents. Unlike traditional surveys, it may be determined only at the end of an ethnographic study. Random sampling justifies the use of survey weighting techniques to estimate characteristics of the sampling frame from the sample. However, generalization beyond the sampling frame to a larger population is especially problematical when the population is not well characterized. An example of incomplete snowball sampling is weighted two ways. Weighting by the reciprocal of the selection probability may be interpreted as projecting each respondent to represent also others in the sampling frame who might have been selected instead. Alternatively, poststratification weights can project each respondent to others linked in the network. Random Samples, Ethnography, Snowball Sampling, Sampling Frames, Sample Weighting.
N. 30, March 1991
THE USE Of CAPI FOR OFFICIAL BRITISH SURVEYS
Nicholas Bateson, Paul Hunter
The introduction of CAPI systems2 seems likely to constitute the largest change in face-to-face interviewing procedures since social surveys began. A brief account is given below of work done up to mid-1990 to develop CAPI systems for surveys carried out on behalf of British government departments by the Office of Population Censuses and Surveys (OPCS). The paper continues with a discussion of the merits and problems of CAPI that have become apparent so far and some of the implications for a survey organisation of its introduction. CAPI, Computer Assisted Personal Interviewing.
USE OF CAPI WITH THE U.S. NATIONAL HEALTH INTERVIEW SURVEY
Owen Thornberry, Benjamin Rowe, Ronald Biggar
The National Health Interview Survey (NHIS) is a continuous cross-sectional survey of the civilian population of the U.S., involving annual interviews with 48,000 households (120,000 persons). Data collection instruments include a core health and demographic questionnaire and one or more supplements designed to address current health issues. For several years, research and developmental work have been pursued via a computer-assisted personal interview (CAPI) system, which, since 1988, has been used with the NHIS to collect data on AIDS (acquired immune deficiency syndrome) knowledge and attitudes: there are plans to rely entirely on the CAPI mode by 1992. A summary is provided of the research and development work to date, and substantive uses of CAPI are described. Issues addressed include: hardware and software choices, data quality, timeliness of data release, quick- turnaround survey capability, and cost. National Health Interview Survey, Computer-Assisted Personal Interview (CAPI), AIDS, Data Quality, Data Collection.
THE COHORT IN TIME AND SPACE: CONCEPTUAL ISSUES AND PRACTICAL CONSIDERATIONS
The increasing emphasis in social research on the life course and the consequent methodological emphasis on longitudinal studies have led to the need for exploration and clarification of the concepts which underlie such studies. One of these crucial underlying concepts is that of time. The article explores the ways in which our assumptions about time impinge on a longitudinal cohort study, using the Scottish Young People’s Survey as an example. Following and developing the themes of Hareven (1982), we can identify the time contexts of individuals, families and other social institutions as being within different frameworks and on different scales. A cohort study, following individuals in time, loses any sense of the time frameworks of other family members. This leads, inter alia, to false assumptions about the homogeneity of families of origin and about the effects of family variables on individuals. Examples are given. The article considers the implications for data collection of background information and analysis which uses background variables. The article concludes by asking how longitudinal can a longitudinal cohort study be, when the emphasis is on the individual and the study is unable to assimilate the time context of the social institutions in which individual cohort members live and with which they negotiate their own life courses. Longitudinal cohorts, life course.
N. 31, June 1991
METHODOLOGICAL, STATISTICAL AND PRACTICAL ISSUES ARISING FROM THE COLLECTION AND ANALYSIS OF WORK HISTORY INFORMATION BY SURVEY TECHNIQUES
This paper examines the methodological, statistical and practical problems encountered in collecting and analysing work and life history information. For illustrative purposes, it draws upon recent experience of and data from the United Kingdom Social Change and Economic Life Initiative, a major socio-economic research programme sponsored by the Economic and Social Research Council and conducted between 1985 and 89. Additionally, it reviews survey-based approaches to the collection of life and work history information since the mid 1970s, and looks at forthcoming applications of the technique in national surveys. The paper divides into three major sections. Part One examines the rationale for collecting life and work history information. Part Two considers the methodology of data collection, the practical problems likely to be encountered and the techniques for tackling such problems. This section examines recent major surveys conducted within the United Kingdom which have collected life and work history information, together with the prospect for further developments in this area. Part Three is concerned with data management and statistical analysis, specifically focussing upon areas of difficulty which are particular to life and work history data. Life Histories, Work Histories, Surveys, Data Collection, United Kingdom.
SIX ANNEES D’ENQUETE SUR LES HOMO- ET BISEXUELS MASCULINS FACE AU SIDA: LIVRE DES DONNEES
Michael Pollak, Marie-Ange Schiltz
Six Years of Surveys on Male Homo- and Bisexuals in Confrontation with AIDS – The Data Report. A criticism of statistical constructions is all the more necessary when the object of study is a population difficult to delimit and hard to encounter (male homosexuals) and when it is a question of such intimate details as sexual practices in relation to a stigmatizing epidemic. With repetition of a survey the question of compatibility of results arises, given the fact that each year the sample is different. Along with this there are the problems of the effect of 6 years of time between the first and last survey (effects of age and change of generations). Also, the questionnaire has changed over these 6 years. The first in 1985 was a research questionnaire whose exploratory character explains the presence of open questions that were replaced later by closed questions. However, the comparison of open and closed responses is informative. On the other hand, closed binary questions were increased and the effect of their increased on the resulted was studied. Multiple verifications of the sample, of questionnaire coherence, and question formulation permits us to judge the robustness of our research instrument. The stability of the social-demographic structure of the respondents over the years permits us to follow the evolution of changes engendered by the epidemic. Sexual Behavior, Male Homosexuals, Surveys, Questionnaires, AIDS.
THE RELATIONSHIP BETWEEN MODE OF ADMINISTRATION AND QUALITY OF DATA IN SURVEY RESEARCH (FINAL PART)
Johannes van der Zouwen, Edith D. de Leeuw
In a series of meta-analyses on the outcomes of ‘mode experiments’, we found small, but statistically significant, and consistent, effects of the method of data collection (‘mode effects’) on the quality (i.e., the representativity, completeness, and accuracy) of the data gathered in surveys. On several indicators of data quality, the face-to-face interview scores slightly better than the telephone interview, but these differences are gradually disappearing over time. Mail surveys usually have lower response rates and less item response than face-to-face interviews and telephone interviews. Mail surveys result in more accurate, less biased, answers than both forms of interviewing, especially when sensitive or embarrassing questions are being asked. In order to give an explanation for these mode effects, we distinguished three ‘mode characteristics’, or intervening variables, (the ‘persuasion power’ of a mode, the complexity of the task for the respondent, and the degree of control over the question-answer process, i.e., control over role-related behavior of respondents and interviewers) and one conditional variable (the degree to which the response categories of the questions differ with respect to their social desirability). These variables were ‘inserted’ in a model which has as independent variables the three modes concerned, and as dependent variables the response rate, and the completeness and accuracy of the responses. Predictions about mode effects on data quality, derived from this model, were confronted with data from a recent mode comparison held in the Netherlands, in which questions about (a.o.) feelings of loneliness, self-esteem, income, and well-being, were posed in face-to-face interviews (N=243), telephone interviews (N=265), and in a mail survey (N=254). This confrontation resulted in a further refinement of the explanatory model, which in its turn may help survey researchers to optimally choose and implement the method of data-collection. Face-to-face Interviews, Telephone Interviews, Mail Surveys, Data Quality, Mode Effects, Meta-Analysis.
DUALISME ET PLURALISME EN SOCIOLOGIE
Jean Michel Berthelot
Dualism and Pluralism in Sociology. This text examines the epistemological meaning of the multiple dualist oppositions regularly employed in sociology, examining them in the light of the concept of incommensurability introduced by the philosophy of science. The advantages and the limits of the use of these oppositions are discussed: even if sociological analysis employs techniques and methods that are often incompatible with the meta-discourse, in concrete argumentation these techniques and methods to be used in a variety of manners. The author proposes the concept of an interlanguage to describe these mechanisms of logic which allow such combinations. Dualism, Pluralism, Methodology, Incommensurability, Meta-Discourse, Interlanguage.
N. 32, September 1991
IDEAL TYPES OR EMPIRICAL TYPSE: THE CASE OF MAX WEBER’S EMPIRICAL RESEARCH
In methodological discussion the concept of the ideal type is closely related to the name of Max Weber. This paper discusses Weber’s own empirical research at the beginning of the century and his application of the term “type” in this research. It seems that his description of the ideal type as a “researcher’s construction”, as “utopia” constructed by composing and enhancing certain aspects of social reality, as an “artificial construction” is indeed the opposite of empirical types. This paper shows that Weber’s articles, which were written in close relationship to his empirical research between 1905 and 1912, use the term type in a different manner. Ideal Type, Empirical Type, Max Weber, Classification Analysis.
DESIGN ISSUES IN THE BRITISH HOUSEHOLD PANEL STUDY
David Rose, Nick Buck, Louise Corti
The British Household Panel Study is the largest single project ever funded by the UK Economic and Social Research Council. Housed in the ESRC’s Centre on Micro-social Change in Britain at the University of Essex, the BHPS will comprise an initial 5000 households and 10000 individuals. This paper introduces the Study through an overview of its coverage and some key aspects of its design. A short discussion of some of the analytic advantages of panel data is complemented by three examples of substantive research issues where panel data can assist in our understanding of micro-social change. The BHPS is discussed as an instance of a dynamic approach to social change and its overall rationale is briefly examined. The Study’s six substantive research areas – household organisation and dynamics; labour market behaviour and activity; income and wealth dynamics; housing; health; and socio-economic values – are described. The remainder of the paper is concerned with three key design issues for panel studies and explains how these have been tackled on the BHPS. The issues are (1) changing populations; (2) non-sampling errors (and especially non-response and panel conditioning); and (3) sample design. The paper is accompanied by a substantial bibliography on panel design and methodology. Micro-Social Change. Panel Studies. Following Rules. Nonresponse. Panel Conditioning. Longitudinal Sample Design.
ONTOLOGICAL AND EPISTEMOLOGICAL PRESUPPOSITIONS OF SOCIAL THEORY
K. Walter Schwager
Empiricist sociology is based on distinct epistemological and ontological presuppositions which usually remain unacknowledged and unscrutinized. This “empiricist metatheory” assumes that language fits reality, i.e., that all descriptive concepts match characteristics in reality; therefore all descriptive concepts can be admitted into scientific theory. These concepts are supposed to be understood, shared, isolated, and static. Measurement now has to reflect the logically prior meaning of these concepts, but measurement remains inherently partial and imperfect. Theory is seen as the probabilistic network of relationships linking the presumed “true values” on these concepts. However, the history of science demonstrates that concepts and their measurement procedures change over time, in a search for better and stronger theories. The assumptions of empiricist metatheory are therefore erroneous, given the history of science, and its methodology removes the impetus towards conceptual change and measurement modification which underlies progress in the natural sciences. Instead concepts and their measurement procedures should be seen primarily as tools for theory development, and as changeable in the pursuit of better theories. Social Theory, Empiricist Sociology.
N. 33, December 1991
PERSUASION STRATEGIES FOR REDUCING REFUSAL RATES IN TELEPHONE SURVEYS
Johannes H. Smit, Wil Dijkstra
The verbal interactions between interviewer and respondent of the introductory part of 85 telephone interviews were audiotaped and analyzed to investigate whether the respondent’s refusal or participation could be ascribed to differences in interviewing behavior. Content analysis showed that, among other things, refusing behavior appeared to depend on the kind of the requests made by the interviewer, and the amount of information provided by the interviewer. Telephone surveys, Refusal rates.
MULTIMETHOD ANALysis: OFFICIAL BIOGRAPHIES OF MEMBERS OF THE CENTRAL COMMITTEE OF THE SOVIET UNION COMMUNIST PARTY
Karl M. van Meter, Lise Mounier, Ghislaine Chartron, Max Reinert
Using official biographies of the 503 members of the Soviet Central Committee between 1980 et 1987, we have previously published the results of an analysis of these biographies for occurrences of geographical names using factorial correspondence analysis. In this study, we use two methods of analysis: LEXINET followed by both a factorial analysis and a typological classification analysis; and ASCESTE, a hierarchically descending classification analysis, both applied to all words and dates occurring in these biographies. The results of these two approaches are compared to verify the stability of the results and their independence in relation to the method that produced them. Soviet Union, Biographies, Factorial Analysis, Classification Analysis, Content Analysis.
SERIATIONS EN BLOCS, MODALITES EN GRAPHIEU BERTIN – UTILISATION ET CONCEPTION LOGICIEL EN GEO-LINQUISTIQUE
Joël Le Fourn, Alain Mailles
Seriation in Blocks – Bertin Graph Modalities – Use and conception of a software in geo-lingistics: Data processing by seriations block methods is based on maintaining the differentiation of individuals observed. The format of the data allows a rich interpretation of the relationship between individuals and the partitioning criteria. In Bertin Gaphis, the suitable tool is orderable matrix, whose foundations are a consequence of the rules of Graphic Semiology. The example of geo-linguistic concerning the personal pronoun subject in Forez allows an interactive way to go from the the matrix to the map, which is the basic idea of the TMC software, which can be regarded as a spreadsheet with mapping capabilities. Bertin Graphs, Seriation in Blocks, Geo-linguistics, Cartography.
N. 34, March 1992
GRADUATE RESEARCH TRAINING IN A NUMBER OF EUROPEAN COUNTRIES AND THE UNITED STATES
Following a survey in the countries concerned, carried out at the request of the Dutch government, this study presents the situation of graduate (“third cycle”) studies and research in several European countries (Netherlands, France, West Germany, Belgium, Denmark, Italy, Spain, United Kingdom) and in the United States. A comparison is made between the American system and the European systems, and suggestions for the future are also made. The conclusion of the international committee involved in this study is included as an annex. Education, Research, Graduate Studies, Europe, United States.
LA MESURE DE L’HETEROGAMIE, DE LA MOBILITE SOCIALE ET DE LEUR RELATION EN FRANCE: LA CONSTRUCTION SECONDAIRE DE DEUX VARIABLES A PARTIR DE L’ENQUETE “EMPLOI 1989” DE L’INSEE
Measurement of Heterogamy, Social Mobility and their Relationship in France: Secondary Construction of Two Variables Based on the 1989 INSEE “Employment” Survey. Based on the INSEE “Emploi” (Employment) survey, a table is constructed which crosses the social origines of male and female members of a couple. The table is then used in the study of social trajectories. Social Mobility.
N. 35, June 1992
EXPLANATIONS FOR DIFFERENCES IN PUBLICATION RATES BETWEEN MALE AND FEMALE ACADEMICS AND BETWEEN PRODUCTIVE AND LESS PRODUCTIVE WOMEN
In this article I will present some possible explanations for the general finding in the literature that men publish more than women and shall present the results of my research in which I compare women with men and productive women with less productive women. In 1973 Cole and Cole found that academic men publish more then academic women. This finding has been replicated many times since then. In later publications it even became a kind of tacit fact that was taken for granted but no longer questioned. Research questions then focussed on phenomena such as why married women publish more than single women and why women who have children do not publish less then women who do not have children (Cole & Zuckerman, 1987; Kyvik, 1990; Thagaard, 1988; Toren, 1989). The question why women publish less then men seems no longer to dominate which implies that this kind of tacit knowledge no longer has the possibility of being corrected. In my research I matched pairs of men and women working at the university of Leiden according to the following variables: the year and field in which they started their university career and the rank to which they were first assigned. In that way the duration of their career and the working environment were held constant. The number of relatively well-matched persons who could be identified from data provided from personnel records made available for this purpose was 198, that is, 99 pairs. Each of these 99 pairs received a questionnaire concerning their working conditions, their publications, their career and private lives. In total 112 questionnaires were returned, 72 from women and 40 from men. Of these, 66 questionnaires were from 33 matched pairs. I compared the publication rates of men and women and made the hypothesis that in the same working environment women do not publish less than men. Furthermore, I compared women who had published much and women who had published little. My hypothesis is that productive women have significantly more time to publish and better working conditions at work and at home. Publication rates.
ANALYSIS OF SEXUAL BEHAVIOR IN FRANCE (ACSF) : WHAT KIND OF ADVANCE LETTER INCREASES THE ACCEPTANCE RATE IN A TELEPHONE SURVEY BEHAVIOR ?
ASCF Group (coordinators Nathalie Bajos, Alfred Spira)
In December 1990, a pilot survey was carried out to measure the effect of an advance letter on the acceptance rate in a telephone survey in France on sexual behavior. The results of this pilot study showed that the acceptance rate could be increased by sending to the household an advance letter announcing a survey on health and prevention a few days before the survey began. The acceptance rate of the survey increased from 69% when no letter was sent to 81% in the case of an advance letter (p<0.001). For legal reasons, another pilot survey had to be carried out in June 1991 to test a modified advance letter announcing a survey on sexual behavior and Aids. The results of this second pilot survey showed an acceptance rate of 54%. So, it appears that an advance letter is useful only if it is judiciously written. Otherwise, it may have a negative effect on the acceptance rate. Sexual behavior, Survey, Telephone survey, Acceptance rate, Comparison of mode of investigations, Advance letter.
Quelles contrées intéressent les gens d’ici? Une méthodologie de traitement pour les tableaux de contingence bi-spatialisés
Alain Mailles, Jean-Pierre Montalieu
Which regions interest the locals? A methodology for bi-spatial contingency array processing: The methodology presented here was carried out to attempt the construction of regionalizations of France withn the framework of research on a geography of spatial representations. Car-window stickers quoting places are the concrete basis for the observations. From the set of more than fifty thousand observations distributed over the entire territory, we constructed a contingency array of 95 rows and 95 columns, each of the two margins being directly associated with a map of France. The following processing methods were used: analysis of multivariate data followed by a cartography, reorderable matrix interactif processng, synthesis mapping. They provide the basis of a methodology capable of providing the basis for the construction of softwar. Bertin Graphics, Data Analysis, Block Seriation, Representational Geography, Computer cartography.
N. 36, September 1992
HIV INFECTION IN AN URBAN SOCIAL NETWORK: A PROGRESS REPORT
A. S. Klovdahl, J. Potterat, D. Woodhouse, J. Muth, S. Muth, W. W. Darrow
The completion of the second year of the Colorado Springs Study provided new insights into the location of HIV infection in a large urban social network. About 250 persons were interviewed by the end of the second year and provided information on over 3500 reported social relationships. Roughly 2000 persons were found to be part of a core connected region which included six individuals confirmed HIV positive. The density of social ties in this core region was about 0.01. The average number of steps (along shortest paths) between HIV-infected persons and others in connected core of this large urban social network was quite small (between 4 and 6). Some implications of the observations are discussed. When individuals are connected together to form large social networks, the concepts and methods of network analysis can lead to a better understanding of factors affecting the spread of infectious agents transmitted in the course of close or intimate personal contact. A better understanding of the factors involved, in turn, can lead to more effective disease control strategies (Klovdahl, 1985). Social Networks, AIDS, Shortest Paths, Urban Communities, Infection.
ESTIMATION ON THE BASIS OF SNOWBALL SAMPLES: HOW TO WEIGHT?
Tom A. B. Snijders
Paper presented at the Workshop on Generalizability Questions for Snowball Sampling and Other Ascending Methodologies, Groningen, 20-21 February, 1992. What are the possibilities of snowball sampling, if one desires valid statistical inference without making probabilistic assumptions on the network structure? In a critical review of the possibilities of snowball sampling for a population of vertices connected by a network of arcs, it is argued that the snowball method is much more suitable for the estimation of parameters of the network structure (or parameters of the population of arcs) than to estimate parameters of the population of vertices. Further work needs to be done to relax the assumption of randomness of the initial sample of the snowball. Snowball Sampling, Weighting, Parameter Estimations, Social Networks.
RARE POPULATIONS, HIDDEN POPULATIONS, AND LINK-TRACING DESIGNS: WHAT AND WHY?
Paper presented at the Workshop on Generalizability Questions for Snowball Literature on snowball sampling and other ascending methodologies (intensive data collection methodologies) is widely scattered. In order to indicate a structure in the diversity of articles, snowball sampling and other ascending methodologies are embedded in the general concept of link-tracing methodologies. From this point of view link-tracing methodologies are considered as tools for analysing social structure and for locating members of special populations. By introducing a new theoretical concept for defining special populations, a subdivision is made in link-tracing methodologies as a tool to locate rare populations and as a tool to locate hidden populations. Hidden Populations, Rare Populations, Ascending Methodologies, Link-Tracing Methodologies, Social Network Analysis.
N. 37, December 1992
LA RESURRECTION DES CLIQUES: APPLICATION DU TREILLIS DE GALOIS
The Ressurrection of Cliques: Application of Galois Lattices. The mathematical definition of clique has never been entirely satisfactory when it comes to providing a procedure for assigning individuals to groups. This paper shows how the clique definition, when it is linked with Galois lattices, can be used to produce an intuitively appealing characterization of groups – one that is consistent with ethnographic descriptions. Two examples, using “classical” social network data sets, are provided. Social network analysis, cliques, Galois lattices.
GENERALIZATION AND STATISTICAL INFERENCE FROM SNOWBALL SAMPLES
Warren D. TenHouten
Ability to extrapolate from snowball samples depends on the way zero-stage subsamples are obtained. If the zero-stage sampling is done on the grounds of convenience or is guided by ethnographic criteria, then statistical inferences cannot be drawn. However, generalization is possible in three ways: (i) as an ethnographically obtained dataset, the results of a snowball sample can be studied as any other field research dataset; (ii) a snowball sample can usually be made more representative of a population by Schmeidler-type weighting schemes; (iii) by regarding sites as subjects and snowball samples as repeated measures of the sites’ informal social networks, generalization through replication becomes a possible strategy. Site sampling is suggested as one approach to accessing sites; for example, by using as sites profiles over facets with attribute pairs day/night, inside/outside, and commercial/residential. If zero-stage sampling is random, statistical inferences can be made from at least the first two stages of the population of individual network members. Snowball sampling is potentially the sociologist’s data acquisition methodology per excellentium, as it provides for inferences about the population of relationships between network members. Since multiple criteria are used to define targeted populations, and since social relationships are multi-level, a generalization of snowball sampling, Multi-Criterion Snowball Sampling (MCSS), is proposed. MCSS is then applied to the measure of hierarchical and solidarity social roles. Social Networks, Snowball Sampling, Site Sampling, Extrapolation, Nulti-Criterion Snowball Sampling.
ACTEURS, CIBLES ET LEVIERS: ANALYSE FACTORIELLE DES RELATIONS DE CONTROLE INDIRECT DANS UNE FIRME AMERICAINE D’AVOCATS D’AFFAIRES
Emmanuel Lazega, Stéphane Vari
Actors, Targets, and Levers: Factor Analyzing Indirect Control Relationships in an American Law Firm. This paper analyzes Krackhardt (1986) type of network data collected from 36 partners in an American law firm. The data describes strategies of indirect control. A strategy is defined as a choice by an actor (the person who delegates the task) of a lever (the person who is asked to perform the task) who takes care of a target (the person who is to be influenced). Despite the heavy-handedness of this approach, factor analyzing this data provides a simple way to represent actors, targets and levers in the same space. It is also useful to describe a division of control work among peers, as well as roles shaping this form of relational influence. Using the results of this analysis, we classify strategies based on whether actors make use of universal and generalist levers, or of local and specialized ones. Social Networks, Factor Analysis, Social Control, Actors, Levers, Targets.
N. 38, March 1993
STRATEGIES OF NUCLEUS FORMATION IN AGGLOMERATIVE CLUSTERING TECHNIQUES
Kenneth D. Bailey
Clustering techniques have gained rather wide acceptance in both the social and biological sciences. There are a number of popular agglomerative clustering techniques, based upon various clustering algorithms. Unfortunately, there has been little recent attention to the manner of nucleus formation for agglomerative methods. This is an important issue, as the choice of initial nucleus can greatly affect the finished clusters regardless of the subsequent clustering algorithm, and further, can have differing effects for different algorithms. This paper examines the effects of four techniques of nucleus formation, an objective and a subjective method of minimizing within-cluster variance; and an objective and subjective method for maximizing between-cluster variance. I conclude that the standard practice of initiating the cluster with a reciprocal pair is recommended for “within” cluster analysis for both single-linkage and complete-linkage methods. However, for “between” clustering, subjective nucleus formation is often preferred for single-linkage analysis. “Between” nucleation methods, either objective or subjective, prove unsatisfactory for complete-linkage clustering, as they tend to select nuclei which are not sufficiently central for cluster formation, and may even be outliers. Cluster Analysis, Nucleus Formation, Clustering Algorithms, Single-linkage, Complete-linkage.
THE USE OF CAPTURE-RECAPTURE METHODOLOGY TO ESTIMATE AND DESCRIBE COVERT POPULATIONS: AN APPLICATION TO FEMALE STREET-WORKING PROSTITUTION IN BLASGOW
Alastair Leyland, Marina Barnard, Neil McKeganey
Study objective – The aim was to estimate the size of the female street-working prostitute population in Glasgow and thereby, in conjunction with collected HIV seroprevalence data, estimate the prevalence and exposure of HIV infection in this population. Design – Street-level interviews were conducted using a short, non-standardised format on 53 nights covering a period of 7 months. A unique identifier (consisting of initials and date of birth) was obtained from each woman each time she was seen working. Saliva samples obtained from the women were analysed for presence of antibodies to HIV. Setting – The major “red light” area in Glasgow, Scotland. Participants – 206 female street-working prostitutes; saliva samples were requested from 197 (96%) of the women and 159 (81%) of these provided specimens. Main results – The prevalence of HIV infection of 2.5% (95% confidence interval 0.7% – 6.3%; n=159) may be combined with a point estimate of the population size over 12 months of 1147 to suggest that approximately 29 HIV positive women may work on the streets of Glasgow in the course of a year. The population may be subdivided into a group of 30 “permanent” workers and a larger group who may cease working and are replaced in equal numbers at a rate of approximately 19 per week. Conclusions – It is possible not only to estimate the size of a covert population but also, by the collection and analysis of longitudinal data, to identify subgroups of workers whose working patterns differ from the remainder of the population. Prostitutes, Street-level Interviews, HIV Infection, Covert Populations, Longitudinal Data, Capture-recapture.
A NEW APPROACH TO POVERTY DYNAMICS
Karl Ashworth, Robert Walker
Studies of poverty that exploit panel data have typically been constrained by the need to address the problem of spell censorship (censorship occurs when spells of poverty begin or end outside the observation period provided by the panel study). The technical solution – a creative blend of simulation and event-history analysis – has had a number of undesirable consequences. It has shifted the unit of analysis from individuals to spells and so inhibited study of the incidence or prevalence of poverty. Attention has focused on the duration of poverty at the expense of severity. It has also proved impossible to examine the pattern of poverty experienced by individuals over time. This paper reports work, based on life course analysis, which attempts to overcome these deficiencies and to exploit the richness of panel data in order better to differentiate the experience of poverty. The data relate to the U.S. Panel Study of Income Dynamics. Poverty, Panel Data, Spell Censorship, Life Course Analysis.
N. 39, June 1993
MIGRATION AND THE LIFE COURSE
Migration events during life course describe a biographic process which is part of the individual biography. Unlike other demographic processes the migration process can be described by a lot of various factors like the frequency of moves, the age when changing residence, the distance, the direction, the motives. In West Germany all these factors have changed for the birth cohorts after the Second World War. Demographic events during the life course are based on experience and commitments in the past life span and on individual goals and ambitions in the future. Therefore, each change of residence has both an retrospective and prospective aspect. Additionally, changes in other biographic processes like marriage or births of children in the familial career or changes in the occupational career directly influence migration decisions. But there are also effects in the opposite direction: the migration process can influence the further development of other biographic processes. Thus, there exist a very complex structure of strong interdependent relationships between the several biographic careers. Migration, Demographic Processes, Life Course.
THE EVALUATION OF MEASUREMENT INSTRUMENTS BY META-ANALYSIS OF MULTITRAIT-MULTIMETHOD STUDIES
Annette Scherpenzeel, Willem Saris
The quality of survey data can be affected by many characteristics of measurement instruments, like the length of the question and the introduction to the question, the form of the response scale, the number and labelling of response categories, the position and context of a question and the data collection technique. An approach that is often used to evaluate measurement instruments is the multitrait-multimethod design. This approach was first introduced by Campbell and Fiske (1959), who suggested to measure each of a number of traits with a number of different methods, and provided guidelines to infer convergent and discriminant validity directly from the multitrait-multimethod correlation matrix. In recent work, confirmatory factor analysis is used mostly to analyse multitrait-multimethod data and to estimate validity, reliability, and method effects. Andrews (1984) first used this approach in a large scale study to evaluate many different measurement instruments across American and Canadian surveys. Very recently, a modified version of the causal model used to analyse multitrait-multimethod matrices with confirmatory factor analysis has been developed. This second order factor model is used in an international comparative research project that is carried out at the moment. In this international project, 15 countries are collecting multitrait-multimethod data, and a meta-analysis will be carried out across all these datasets, to get an evaluation of measurement instruments in the line of Andrews’ study. The procedure that is followed in the international project is demonstrated by a large scale study of life satisfaction. For this study, data about satisfaction with life in general and satisfaction with different aspects of life were collected in 10 different european countries. It is shown how the information that results from the meta-analysis across these datasets can be used to correct for measurement error in a substantive model of life satisfaction. Measurement Instruments, Secondary Analyse, Multitrait-multimethod Studies, Life Satisfaction, International Comparative Research.
EFFECTS OF MEASURES AIMES AT INCREASING THE QUALITY OF RECALL DATA
Johannes van der Zouwen, Wil Dijkstra, Wander van der Vaart
In a field experiment the effects of four different measures, aimed at increasing the data quality of retrospective questions, were assessed: 1) a ‘checklist’ for factual questions, 2) use of backward recall – instead of the usual forward recall – for the reporting of a series of life-events, 3) a ‘cue-list’ for attitudinal questions, and 4) the use of a ‘time-line’ for the reconstruction of sequences of particular life-events. Only the first and last mentioned measures showed some of their assumed beneficial effects. The results of our analysis indicate that quality of information collected with retrospective questions is not very good and that the measures aimed at improving data quality are generally not very effective. This means that life history research, largely based on answers to retrospective questions posed in survey-interviews, is a risky affair. Recall Data, Retrospective Questions, Recall Error, Check-list, Cue-list, Time-line, Backward Recall.
<END OF FILE>