Multivariate Multinomial Logit Models with Associations among Dependent Variables

Abstract

The authors introduce a new group of multinomial logit models with special contrasts to identify covariate effects on multiple categorical dependent variables that are strongly associated with each other. The authors first develop the method for a case with two dependent variables and then extend the method to a case with three dependent variables. The model can account for both nominal and ordinal scales of categorical dependent variables. The authors formulate the covariate effects to represent unique effects on each dependent variable so that they become independent across different dependent variables. The application focuses on the multiplicity of occupational attainments by analyzing how gender, race, educational attainment, and parental occupation characteristics affect three distinct but nonindependent dimensions of occupations: socioeconomic status, social skill level, and math and science skill levels.

Keywords

multinomial logit models multiple categorical dependent variables occupational attainment

Get full access to this article

View all access options for this article.

References

Agresti

Alan

. 2010. Analysis of Ordinal Categorical Data. 2nd ed. Hoboken, NJ: John Wiley.

Clogg

Clifford C.

1982. “Some Models for the Analysis of Association in Multiway Cross-Classifications Having Ordered Categories.” Journal of the American Statistical Association 77:803–15.

England

Paula

. 1992. Comparable Worth: Theory and Evidence. New York: Aldine.

Goodman

Leo A.

1979. “Simple Models for the Analysis of Association in Cross Classifications Having Ordered Categories.” Journal of the American Statistical Association 74:537–52.

Shrout

Patrick

Kandel

Denise B.

1981. “Analyzing the Property of the Dyads: Determinants of Similarity in Marijuana Use in Adolescent Friendship Dyads.” Sociological Methods and Research 9:363.

Sun

Bingrui.

2014. “Bivariate Multinomial Logit Models.” Doctoral dissertation, Department of Mathematics and Statistics, Memorial University of Newfoundland, St. John’s, Newfoundland, Canada.

Vermunt

Jeroen K.

1997a. lem: A General Program for the Analysis of Categorical Data. Tilberg, the Netherlands: Tilberg University.

Vermunt

Jeroen K.

1997b. Loglinear Event History Analysis. Newbury Park, CA: Sage.

Yamaguchi

Kazuo

. 2013. “New Regression Models with Egocentric Social Network Data: An Analysis of Political Party Preference in Japan.” Sociological Methodology 42:257–85.