Using Replication Methods for Variance Estimation in Proportional Odds Models for Complex Sample Data

• In Event: Division D: Multiple Linear Regression: The General Linear Model SIG

Abstract

When conducting the ordinal regression analysis, we assume that the data are collected with simple random sampling. However, more complex sampling designs are often used for large-scale national studies. Due to confidentiality and privacy, public-use data such as the HSLS:09 and PISA data often do not release the information of PSUs and strata to researchers. Thus using the replicated sampling methods is the only option. The purpose of this study is to apply the replication methods for variance estimation in PO models to complex sample data in educational research. In addition, it compares the results of variance estimations from different types of replication methods for PO models.

Authors

• Xing Liu, Eastern Connecticut State University

• Haiyan Bai, University of Central Florida

• Hari P. Koirala, Eastern Connecticut State University