No full-text available
To read the full-text of this research,
you can request a copy directly from the author.
La credibilite des reponses dans les etudes sur des themes controverses : Aspects pratiques de l'analyse des etudes sur la consommation de drogues
Abstract
Ce papier présente la problématique liée aux études sur des thèmes controversés tels que le comportement sexuel, l'avortement ou la consommation de drogues. A partir d'une étude empirique à grande échelle (4 384 enquêtes) sur la consommation de drogues, il montre les principales difficultés de l'analyse dans ce domaine. Il présente aussi diverses techniques permettant de prouver la validité de ce type de recherches et améliorant la compréhension des résultats obtenus.
ResearchGate has not been able to resolve any citations for this publication.
The quality of drug data in the 1984 wave of the National Longitudinal Survey of Youth is explored. Comparisons with other national surveys indicate that underreporting of use of illicit drugs other than marijuana appears to have taken place, and that light users of these drugs are underrepresented among the self-acknowledged users. Comparison with marijuana use reported four years earlier indicates that experimental marijuana users are much less likely than extensive users to acknowledge involvement. Even after controlling for frequency of use, underreporting is more common among terminal high school dropouts and minorities. Not only individual characteristics but field conditions also contribute to underreporting. Familiarity with the interviewer, as measured by number of prior interviewing contacts, depresses drug use reporting. We speculate that interviewer familiarity increases salience of normative standards and that participants respond not only in terms of their past familiarity but also in terms of their subjective expectations regarding the probability of a future encounter with the interviewer.
Methodological studies on mail surveys typically assess the effectiveness of a device with respect to only one dependent variable—response rate. Other factors in the use of mail surveys, i.e. response speed and quality, also require methodological research. This report includes a discussion of these factors and a review of the findings of studies that have included them as dependent variables.
This article presents the results of two split-panel studies, one local and one national, which attempt to measure the effect of four questionnaire variants on the reporting of income in sample surveys. The results tend to support the hypothesis that the more summary the treatment of subject, the less complete is the reporting of income. The observed differences in the results from the different variants, however, were not as large as was anticipated.
This article reports a nationwide study of response effects based on answers from almost 1,200 respondents to threatening behavioral questions presented in various formats. Results indicate that threatening questions requiring yes-or-no answers can be asked in any format, but that threatening questions requiring quantified answers are best asked in open-ended, long questions with respondent-familiar wording.
The basic theories and measurement procedures for reliability and the closely related concept of generalizability are reviewed, illustrated, and evaluated for use in marketing research. A critique is given of a subset of previous marketing research studies in which reliability estimates were used and recommendations are made for future research.
Increasing the amount of family planning in less-developed countries is crucial to their economic development and is basically a marketing job. Much important marketing research has been done in this area; its history is described here. But much more needs to be done.
A critical element in the evolution of a fundamental body of knowledge in marketing, as well as for improved marketing practice, is the development of better measures of the variables with which marketers work. In this article an approach is outlined by which this goal can be achieved and portions of the approach are illustrated in terms of a job satisfaction measure.
The problem of developing reliable and valid measures is potentially greater for drug use than for other behaviors because the use of many of the drugs is Illegal and disapproved of by society. In the absence of independent criteria, responses to drug use questions may be tested for consistency and validity. In a survey based on a representative sample of 8,206 New York State public secondary school students, we find that self-reported Illicit drug use is consistent at one point in time, but less so over time. Self-reported Illicit drug use is strongly related to adolescents' attributes as well as to data independently obtained from best school friends. Only a very small proportion of respondents report the use of a fictitious drug. However, inconsistencies over time are related to sporadic patterns of drug use, suggesting that the failure to report certain types of drug use is more the result of poor recall than of willful concealment of use. Underreporting is potentially a greater problem in drug research than overreporting.
L'efficacité d'un instrument d'sétude peut être appréhendée à partir de trois axes: méthodologique, opérationnel et décisionnaire. Huit échelles de mesure courantes sont ensuite évaluées. l'échelle à supports sémantiques présente une efficacité globale satisfaisante; en revanche, l'échelle de Stapel et, dans une moindre mesure, l'échelle dichotomique sont plutôt faibles. Trois suggestions sont faites pour la construction future d'instruments: simplicité, multivalidation, analyse coût/Performance.
A components of variance model for categorical data from unbalanced designs which is directly analogous to a one-way random effects ANOVA model for quantitative data is proposed. The variance components provide separate reliability measures for each of the response categories and disagreement measures between pairs of response categories in terms of (within subject) intraclass and interclass correlation coefficients. The estimation procedures involve usual MANOVA calculations which can be expressed as compounded functions of the multinomial observations. Thus, the variances of these estimates can be obtained from linearized Taylor series results. These procedures are illustrated with data from a psychiatric diagnosis study.
This article considers the feasibility of measuring responding desirably with attitude-opinion items, and presents and index based on 16 such items. Estimates of reliability and validity for this index are given, and examples of its use as a covariate (control) in attitude research are presented. Similarities and differences from other related scales are discussed.
Abstract Perceived normative threat influences responses to questions in surveys. Respondents who report that questions about an activity
would make most people very uneasy are less likely to report ever engaging in that activity than are persons who report less
uneasiness. If respondents do not admit to participating in an activity, perceived threat appears to have acted as a gatekeeper
to prevent further questions. Since perceived threat is associated with underreporting, some simple adjustment methods may
be used to improve behavioral estimates.
En marge des problèmes traditionnels de fidélité et validité de la mesure en marketing, la cohérence interne des réponses apportées par les répondants dans une enquête par interview est étudiée. Confrontés à des questions complexes portant sur des produits très nouveaux, les répondants font preuve d'un certain degré d'incohérence dans leurs réponses. La mesure et les facteurs explicatifs de cette incohérence doivent être examinés pour redresser les biais éventuels. Lorsqu'il pressent un danger d'incohérence sur un sujet, le chercheur en marketing doit impérativement multiplier les formulations des questions relatives à ce sujet.
The ability of the randomized response technique to obtain sensitive information on surveys was compared to two other methods involving self-administered questionnaires. Using a national probability sample, and extending upon a number of developmental changes which have evolved over the years, the randomized response technique proved to be decidedly superior to the other methods.
In follow-up of 1,500 drug-using applicants to the NIMH civil commitment program under Titles I and III of the Narcotic Addict Rehabilitation Act, efforts were made to measure the reliability and validity of self-reported criminal and drug-taking behavior. Various methods to assure reliable and valid responses were developed and implemented. These methods included choice of interviewer, intraquestionnaire safeguards, interview-reinterview procedures, and use of police records and urinalysis reports. Overall, the results indicated that the responses for 829 respondents were reliable. The greatest limitation to the validity study was the incomplete and unreliable police records and urinalysis reports.
Smyllie et al. [1965] and Armitage et al. [1966] have discussed several measures of disagreement between k observers in recording the presence or absence of a clinical sign in n patients. In this paper such measures are examined with respect to their sampling aspects using a multinomial model of the present author (Bennett [1967; 1968]). Generalizations are then considered for treatment of data in which the presence or absence of several signs is recorded.
Testing validity on controversial subjets: the case of drug prevalence in Peru, 42nd Conference of the American Association for Public Opinion Research
- J Jutkowitz
- R Arellano
Cocaine use in America: Epidemiological and clinical perspectives
- N Kozel
- E Adams
Threatening questions for the public in a survey about AIDS, 42nd Conference of the American Association for Public Opinion Research
- D Binson
- P Murphy
- D Keer
Guide to drug abuse research terminology
- J Nelson