Journal article
Case-Deletion Diagnostics for Maximum Likelihood Multipoint Quantitative Trait Locus Linkage Analysis
Human Heredity, Vol.67(4), pp.276-286
03/2009
DOI: 10.1159/000194980
PMCID: PMC2880723
PMID: 19172086
Abstract
Objectives: Case-deletion diagnostic methods are tools that allow identification of influential observations that may affect parameter estimates and model fitting conclusions. The goal of this paper was to develop two case-deletion diagnostics, the exact case deletion (ECD) and the empirical influence function (EIF), for detecting outliers that can affect results of sib-pair maximum likelihood quantitative trait locus (QTL) linkage analysis. Methods:Subroutines to compute the ECD and EIF were incorporated into the maximum likelihood QTL variance estimation components of the linkage analysis program MAPMAKER/SIBS. Performance of the diagnostics was compared in simulation studies that evaluated the proportion of outliers correctly identified (sensitivity), and the proportion of non-outliers correctly identified (specificity). Results: Simulations involving nuclear family data sets with one outlier showed EIF sensitivities approximated ECD sensitivities well for outlier-affected parameters. Sensitivities were high, indicating the outlier was identified a high proportion of the time. Simulations also showed the enormous computational time advantage of the EIF. Diagnostics applied to body mass index in nuclear families detected observations influential on the lod score and model parameter estimates. Conclusions: The EIF is a practical diagnostic tool that has the advantages of high sensitivity and quick computation.
Details
- Title: Subtitle
- Case-Deletion Diagnostics for Maximum Likelihood Multipoint Quantitative Trait Locus Linkage Analysis
- Creators
- Maria C.B MendozaTrudy L BurnsMichael P Jones
- Resource Type
- Journal article
- Publication Details
- Human Heredity, Vol.67(4), pp.276-286
- Publisher
- Basel, Switzerland
- DOI
- 10.1159/000194980
- PMID
- 19172086
- PMCID
- PMC2880723
- ISSN
- 0001-5652
- eISSN
- 1423-0062
- Number of pages
- 11
- Language
- English
- Date published
- 03/2009
- Academic Unit
- Statistics and Actuarial Science; Epidemiology; Biostatistics; Fraternal Order of Eagles Diabetes Research Center; Public Policy Center (Archive)
- Record Identifier
- 9983985811902771
Metrics
20 Record Views