Abstract:
|
Most of regression methodologies are only able to find effects of two-way or three-way interactions. However, in high dimensional data settings, the complex interaction between more than three predictors may present. When all the predictors are binary, Logic regression, a tree-based generalized regression methodology, has been used to construct complex interactions between binary predictors as Boolean logic expressions. This methodology, however, assumes independent observations, and may cause serious inferential problem when directly applied to correlated data, such as repeated measurement and clustered data, which is commonly seen in practice. This paper is going to study the logic regression modeling with correlated data assuming various correlation structures. The proposed method will be compared with the logic regression without considering correlations via simulated data. The impact of different correlation structures will also be studied in simulation studies. The proposed method will be applied to a real data set for syndromic diagnosis of vaginal infections in India and compared to a commonly used algorithm developed by the World health Organization (WHO).
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.