REFERENCES
- Kaestner R. The effect of Illicit Drug Use on the Labor Supply of YoungAdults. J. Hum. Resources 1994; 29: 126–155
- Kandel D. B., Davies M. Labor Force Experiences of a National Sample of Young AdultMen. The Role of Drug Involvement. Youth & Society 1990; 21: 411–445
- Kandel D., Davies M., Karus D., Yamaguchi K. The Consequences in Young Adulthood of Adolescent Drug Involvement. Arch. Gen. Psychiatry 1986; 43: 746–754
- Newcomb M. D., Bentler P. M. Consequences of Adolescent Drug Use. Sage, Newbury Park, CA 1988
- Newcomb M. D., Bentler P. M. Drug Use, Educational Aspirations, and Work Force Involvement:The Transition from Adolescent to Young Adulthood. Am. J. Commun. Psychol. 1986; 14(3)303–321
- Bachman J., Johnston L., O'Malley P. The Monitoringthe Future Project After Twenty-two Years: Design and Procedures. Monitoring the Future Occasional Paper No. 38, Institute for Social Research, Ann Arbor, MI 1996
- Johnston L., O'Malley P., Bachman J. National SurveyResults on Drug Use from the Monitoring the Future study, 1975–1997. Volume I: Secondary school students (NIH PublicationNo. 98-4345), National Institute onDrug Abuse, Rockville, MD 1998
- Thereare several reasons for using unweighted data rather than weighted. (1) Weare not attempting to make population estimates of prevalence (for which weighteddata would be preferable). (2) We are primarily interested in the resultsof multivariate analyses, and the use of weights in multivariate analysesis controversial.9-11(3) The major independent variable of interest is marijuana use, which wasone of the several variables used in stratification of the follow-up sample,and use of weights is such cases can result in a considerable loss of power.[12]
- DuMouchel W., Duncan G. Using Sample Survey Weights in Multiple Regression Analysisof Stratified Samples. J. Am. Statist. Assoc. 1983; 78: 535–543
- Korn E., Graubard B. I. Epidemiologic Studies Utilizing Surveys: Accounting for theSampling Design. Am. J. Pub. Health 1991; 81: 1166–1173
- Pfeffermann D. The Role of Sampling Weights When Modeling Survey Data. Internat. Statist. Rev. 1993; 61(2)317–337
- Winship C., Radbill L. Sampling Weights and Regression Analysis. Sociol. Methods Res. 1994; 23(2)230–257
- Inthe original question there was a fifteenth job category which was labeled“homemaker/housewife”. We excluded this category from the classificationof Table 1 because its “prestige” withrespect to the other job categories is unclear
- Goodman L. A. The Analysis of Dependence in Cross-classifications Having OrderedCategories Using Log-linear Models for Frequencies and Log-linear Models forOdds. Biometrics 1983; 39: 149–160
- The ψjBparameters account for differences in the probabilities of the column variable.Since we do not necessarily expect the population frequencies of the job categoriesto be equal, these parameters have to be present in the model. However, forthe present study they are not of interest. This is similar to regressionanalysis where an intercept is typically required by the model although thesubstantive interest focuses only on the slope coefficient
- Notethat the ψ1A was constrainedto be equal to zero. Therefore, the other ψiAparameters express the difference in the log-odds, log (Fi,j+1/Fij),compared to the non-user group
- Herzog A. R., Bachman J. G. Sex Role Attitudes Among High SchoolSeniors: Views About Work and Family Roles. Institute for Social Research, Ann Arbor, MI 1982