Far East Journal of Theoretical Statistics
Volume 13, Issue 2, Pages 175 - 188
(July 2004)
|
|
ON MISSING PHENOTYPE DATA IN MULTIVARIATE FAMILY BASED ASSOCIATION TESTS: FBAT-GEE-IMP AND IMPUTATION STRATEGIES BASED ON THE EM-ALGORITHM, THE DA-ALGORITHM AND THE CONDITIONAL MEAN MODEL
Amy Murphy (U. S. A.), Kristel Van Steen (U. S. A.) and Christoph Lange (U. S. A.)
|
Abstract: In this paper, we address missingness in multivariate phenotype data in family-based association tests (FBAT-GEEs (Lange et al. [Biostatistics 4 (2003a), 195-206])). We propose a new test statistic, FBAT-GEE-IMP, which is designed for imputed multivariate data. Several imputation methods are discussed, and compared with the original FBAT-GEE, which only uses probands with complete phenotypic data. Using simulation studies to assess the power of the different approaches, we illustrate the practical relevance of the proposed methodology. FBAT-GEE-IMP has been implemented in the software package PBAT and is freely available at http://www.biostat.harvard.edu/~clange/default.htm. |
Keywords and phrases: family-based association tests (FBATs), GEE, multivariate phenotypes. |
|
Number of Downloads: 256 | Number of Views: 790 |
|