simulation genotype data

Chi-Hse Teng teng_chi-hse at REMOVE.lilly.com
Thu Jun 3 09:45:51 EST 1999


Hi,

I have a question about simulating sib pair data.   I appreciate your
help.

If I want to check whether the p-value from linkage analysis package is
"correct" or not, I need to do simulation to generate an empirical
p-value.    When I do the simulation should I keep parents genotype and
just simulate alleles transmission or should I simulate parent's
genotype according to population allele frequencies?    Assume there are
several marker alleles, A, a B, b C, c, D, d, ...

P(LOD=k)= P(LOD=k| parent Aa and Bb)*P(parent Aa and Bb) + P(LOD=k|
parent Cc and Bb)*P(parent Cc and Bb) + many other possible cases.

If we exhaust all the possible cases, we can get the distribution of LOD
score.    When sample size is large, LOD has a Chi-square distribution
or a mixed Chi-square distribution, according to limit theories.

My concern is, when we keep or infer the parent's marker genotype for
simulation, the p-value we get is from the conditional LOD distribution
given the parent's genotype.    Is there a or some parent's genotype
will give us a distribution different from the distribution of LOD?
If there is/are, does it matter?

Thanks.

Chi-Hse



--
Chi-Hse Teng
Please remove "REMOVE." in the reply address before reply.
Thanks.





More information about the Gen-link mailing list