Mixed models incorporating intra-familial correlation through spatial autoregression

Authors

  • George J. Knafl,

    Corresponding author
    1. Yale University, School of Nursing, 100 Church Street South, New Haven, CT 06536-0740
    • Yale University, School of Nursing, 100 Church Street South, New Haven, CT 06536-0740.
    Search for more papers by this author
    • Senior Research Scientist.

  • Kathleen A. Knafl,

    1. Yale University, School of Nursing, 100 Church Street South, New Haven, CT 06536-0740
    Search for more papers by this author
    • Professor.

  • Ruth McCorkle

    1. Yale University, School of Nursing, 100 Church Street South, New Haven, CT 06536-0740
    Search for more papers by this author
    • Florence S. Wald Professor.


Abstract

Family researchers are challenged by the need to account for the special forms of statistical dependence that can exist in family data. To address this issue, mixed modeling methods were adapted to account for dependence of continuous outcomes measured across multiple family members. This was accomplished using a spatial autoregressive approach that accounts for dependence on direction as well as on distance apart. For family data, the dimensions underlying direction can correspond to different family members, thereby accounting for different correlations between family members. When the data are also longitudinal, a dimension representing distance apart in time also can be included to account for temporal correlation. Fixed effects involving general linear models can be included as well. Example analyses were conducted to demonstrate the use of the spatial autoregressive approach for modeling intra-familial correlation. © Wiley Periodicals, Inc. Res Nurs Health 28:348–356, 2005

Ancillary