Improving self-reports of drug-use: numeric estimates as fuzzy sets (2003)

FRANÇAIS :
Les estimations numériques des comportements d'usage tels lâge dinitiation, la durée dune période dabstinence ou les quantités consommées et les fréquences de consommation, font appel chez lusager interrogé à plusieurs réponses sensiblement équivalentes. Dans les questionnaires classiques, quand il est demandé une seule réponse, le choix se porte sur la réponse qui semble la meilleure. La théorie des « fussy sets » suggère que le processus cognitif de réponse à la question conduise à des estimations partiellement valides et teint compte de léventail des réponses qui peuvent être considérées comme vraies. Cette méthode est testée chez 88 usagers de tabac, alcool, marijuana ou amphétamines. Les conclusions montrent que les participants donnent une description plus détaillée et plus complète sur leur usage passé.
ENGLISH :
Aims : This study compared traditional crisp set to fuzzy set representations of quantitative estimates of drug use to investigate what survey respondents recall about their drug use. Design : Sample survey comparing (a) responses to traditional survey questions asking for single numeric estimates to (b) responses to fuzzy set questions asking for boundary estimates of drug use. Setting : Research facilities of a public university in a large southern California city. Participants : Community residents using tobacco, alcohol, marijuana or amphetamine (n=88). Measurements : Personal interviews in which participants reported onset, frequency and daily rate of tobacco, alcohol, marijuana and amphetamine use based on traditional survey questions and fuzzy set questions. Findings : With the exception of marijuana use, fuzzy boundary estimates differed significantly from the traditional point estimates (P < 0.05). Median retest reliabilities for fuzzy boundary estimates were 0.96 (tobacco). 0.81 (alcohol), 0.96 (marijuana) and 0.95 (amphetamine). The degree of vagueness varied by behavior and drug, with estimates of daily rates showing greatest fuzziness and onset of use showing least fuzziness. Conclusions : Participants recall more about past drug use than a single numeric estimate. All numeric estimates of drug use showed fuzzy set properties. Compared to traditional self-reports of drug use, fuzzy set representations provide a more complete and detailed description of what participants recall about past drug use. (Author' s abstract)