Get access

On fitting generalized linear mixed-effects models for binary responses using different statistical packages

Authors

  • Hui Zhang,

    Corresponding author
    1. Department of Biostatistics, St. Jude Children's Research Hospital, Memphis, TN 38105, U.S.A.
    • Department of Biostatistics, St. Jude Children's Research Hospital, Memphis, TN 38105, U.S.A.
    Search for more papers by this author
  • Naiji Lu,

    1. Department of Biostatistics and Computational Biology, University of Rochester, Rochester, NY 14642, U.S.A.
    2. Department of Psychiatry, University of Rochester, Rochester, NY 14642, U.S.A.
    Search for more papers by this author
  • Changyong Feng,

    1. Department of Biostatistics and Computational Biology, University of Rochester, Rochester, NY 14642, U.S.A.
    Search for more papers by this author
  • Sally W. Thurston,

    1. Department of Biostatistics and Computational Biology, University of Rochester, Rochester, NY 14642, U.S.A.
    Search for more papers by this author
  • Yinglin Xia,

    1. Department of Biostatistics and Computational Biology, University of Rochester, Rochester, NY 14642, U.S.A.
    2. Department of Psychiatry, University of Rochester, Rochester, NY 14642, U.S.A.
    Search for more papers by this author
  • Liang Zhu,

    1. Department of Biostatistics, St. Jude Children's Research Hospital, Memphis, TN 38105, U.S.A.
    Search for more papers by this author
  • Xin M. Tu

    1. Department of Biostatistics and Computational Biology, University of Rochester, Rochester, NY 14642, U.S.A.
    2. Department of Psychiatry, University of Rochester, Rochester, NY 14642, U.S.A.
    3. VA Center of Excellence at Canandaigua, Canandaigua VAMC, Canandaigua, NY 14424, U.S.A.
    Search for more papers by this author

Abstract

The generalized linear mixed-effects model (GLMM) is a popular paradigm to extend models for cross-sectional data to a longitudinal setting. When applied to modeling binary responses, different software packages and even different procedures within a package may give quite different results. In this report, we describe the statistical approaches that underlie these different procedures and discuss their strengths and weaknesses when applied to fit correlated binary responses. We then illustrate these considerations by applying these procedures implemented in some popular software packages to simulated and real study data. Our simulation results indicate a lack of reliability for most of the procedures considered, which carries significant implications for applying such popular software packages in practice. Copyright © 2011 John Wiley & Sons, Ltd.

Ancillary