Search Results
Author: Clogg, Clifford C.
Resulting in 1 citation.
| 1. |
Arminger, Gerhard Clogg, Clifford C. |
|
Analysis of Panel Data and Related Types of Data with Binary Outcomes Using Finite-Mixture a New Methods: a New Approach for Studying Unobserved Heterogeneity Presented: Cincinnati, OH, Population Association of America Meetings, April 1993 Cohort(s): NLSY79 Publisher: Population Association of America Keyword(s): Data Quality/Consistency; Heterogeneity; Modeling, Fixed Effects; Modeling, Random Effects; Variables, Independent - Covariate Permission to reprint the abstract has not been received from the publisher. We consider a model for repeated observations of binary outcomes that includes a) covariate effects either fixed or time-varying, b) individual-level heterogeneity that can be regarded either as fixed effects or as random effects, and c) nonparametric modeling of the "unobserved" heterogeneity using mixture methods and scaled latent classes. The model as well as the likelihood theory builds on detailed analysis of a special case presented in Lindsay, Cloff, and Grego (JASA, 1991). We examine the relationship of this framework to other models, such as the Rasch model, and other estimation methods, such as conditional maximum likelihood. The likelihood equations and numerical algorithms for their solution are provided in detail. Examples drawn from a popular longitudinal data set (NLSY) are used to illustrate the flexibility of the approach. The chief benefit of the approach is that the "unobserved" heterogeneity can be "completely" characterized in terms of a set of J+l mixing weights (latent class proportions) and a set of J+1 scores for the latent classes (J is the number of waves or number of outcomes). Using this approach and this model we can estimate the underlying distribution posited for the heterogeneity terms and can compare the distribution estimated under different models (e.g., the model without covariate effects and a model with specified covariate effects). |
|
|
Bibliography Citation
Arminger, Gerhard and Clifford C. Clogg. "Analysis of Panel Data and Related Types of Data with Binary Outcomes Using Finite-Mixture a New Methods: a New Approach for Studying Unobserved Heterogeneity." Presented: Cincinnati, OH, Population Association of America Meetings, April 1993. |