2010년 4회 통계세미나 개최 안내
통계연구소에서는 다음과 같이 통계 세미나를 개최하오니 많은 참여 바랍니다.
일시: 2010년 5월 12일(수) 오후 5시
장소: 고려대학교 정경관 501호
연사: Hwan Chung
(Department of Statistics , Ewha Womans University)
“Latent class profile analysis: an application to stage-sequential process of early-onset drinking behaviors”
Earlier age of drinking is a well-known risk factor for a variety of adverse public health consequences in the United States and worldwide. In longitudinal research on early-onset drinkers, a great deal of attention has been paid to the identification of subgroups of individuals who follow similar sequential patterns of drinking behaviors. However, research on the sequential development of drinking behavior can be challenging in part because it may not be possible to directly observe the particular drinking behavior stage at a given point in time. To address this difficulty, one can use a latent class analysis (LCA) approach, which provides a set of principles for the systematic identification of homogeneous subgroups of individuals. We develop an LCA approach for repeated measures for use in an investigation of stage-sequential patterns of drinking behaviors among early-onset drinkers from early to late adolescence, using data from the ‘National Longitudinal Survey of Youth 1997.’ In this work, an identification procedure is used to characterize different patterns of drinking behaviors into a small number of classes, based on responses to questions about drinking at each measurement occasion; it then examines the class sequencing of early-onset drinkers over the entire set of time points in order to identify two or more homogeneous subgroups. All early-onset drinkers in a subgroup should exhibit a similar sequence of class memberships over time. This approach uncovers four common drinking behaviors in early-onset drinkers over three measurements from 1997 through 2003. The sequences of drinking behaviors can be grouped into three groups of sequential patterns representing the most probable progression of early-onset drinking behaviors. Applying a multinomial logistic regression to the latent responses, we investigate how the sequential patterns of early-onset drinking behaviors might depend upon background factors such as sex, race, and age. Inferences about the model parameters are obtained by a maximum-likelihood method using the EM algorithm.
고려대학교 통계연구소