Measuring the Inflation of the Lod Score Due to Its Maximization
Over Model Parameter Values in Human Linkage Analysis
D.E. Weeks, T.Lehner, E. Squires-Wheeler, C. Kaufmann, J. Ott
Genetic Epidemiology , 7(4), 237--743 (1990)
Abstract
A computer-simulation method is presented for determining and
correcting for the effect of maximizing the lod score over disease
definitions, penetrance values, and perhaps other model parameters.
The method consists of simulating the complete analysis using marker
genotypes randomly generated under the assumption of free
recombination. It is applicable as a "post-treatment" to linkage
analyses of any trait with an uncertain mode of inheritance and/or
disease definition. When the method is applied to a linkage analysis
of schizophrenia versus chromosome 5 markers, we find that, in this
specific case, the P-value associated with a maximum lod score of 3
is equal to 0.0003. We also find that a lod score of 3.0 should be
"deflated" by approximately 0.3 to 1 units, and, by tentative
extrapolation, the observed lod score of 6.5 should be "deflated" by
0.7 to 1.5 units.
