University of Michigan Center for Statistical 
Genetics
Search
 
 

 
 

PSEUDO -- Variance estimation for the replicate pool method

An important feature of the replicate pool method is that the variance structure for the replicate pool p-value estimator (pRP) differs from a standard gene-dropping p-value estimator (pGD). For N simulations, the distribution of pGD will be Binomial(N, p) and

For pRP, the variance will be different because family-specific statistics are reused inducing a correlation among simulated outcomes. Song (2004) studied the variance structure of pRP and showed that

where
When NP is large, this simplifies to

Family variance weights

Family variance weights (Wfc) represent the conditional probability that the replicate sampling for family f on chromosome c drives the outcome of the experiment. PSEUDO estimates this values as

References

Song KK, Weeks DE, Sobel E, Feingold E (2004) Efficient Simulation of P Values for Linkage Analysis. Genetic Epidemiology 24: 1-9

Wigginton JE and Abecasis GR (2005) An Evaluation of the Replicate Pool Method : A Method for Quick Estimation of Genomewide Linkage Peak P-Values. ...


 
 

University of Michigan | School of Public Health | Abecasis Lab