Post by arno on Nov 12, 2013 15:28:51 GMT
Probably another stupid question. I got the following results after running including one GRM (case-control) in REML analysis.
Source Variance SE
V(G) 0.165838 0.016453
V(e) 0.000000 0.015106
Vp 0.165838 0.004259
V(G)/Vp 0.999999 0.091090
The estimate of variance explained on the observed scale is transformed to that on the underlying scale:
(Proportion of cases in the sample = 0.215705; User-specified disease prevalence = 0.030000)
V(G)/Vp_L 1.081173 0.098484
Does anyone understand why the residual variance is 0?
Post by Hong on Nov 13, 2013 5:57:41 GMT
The inflated estimate on the observed scale would be due to some other non-genetic factors. If they genotyped the cases and controls separately, there would be some batch effects.
1) A starting point for pulling apart these results is firstly to look at the QQ plot of the GWAS from these same data and secondly to check that close relatives have been removed.
2) Please note that subtle genotyping errors can be accumulated across the genome and cause inflated estimates even when GWAS signals seem normal.
3) Another test is that one can do an association test according to the batch information that can be inflated if there are batch effects.
4) When there are no such batch effects, one would still expect to see an estimate of h2O being larger than 1 only when h2L is large and K is extremely smaller than P. However, this may not be the case given your example.
Please see the papers below about how to correct for such batch effects.
Estimation and partitioning of polygenic variation captured by common SNPs for Alzheimer's disease, multiple sclerosis and endometriosis. Hum Mol Genet 22: 832-841
A simple and fast two-locus quality control test to detect false positives due to batch effects in genome-wide association studies. Genet Epidemiol (2010) 34: 854-862.