cadeleeuw
commented
1 day ago Hi,
Everything being NA usually happens when the rG estimate is too unstable and therefore too far out of bounds, this would also be accompanied by a warning in the output log.
The p-values and confidence intervals are computed separately in the code, so p-values could indeed fail to be computed even though the confidence intervals have been, though I wouldn't normally expect that to be very common. I would need to know if any LAVA or general R warnings/errors were reported in the output accompanying those, to say more about it.
That said, if the analysis is based on MTAG-generated SNP p-values, that may have something to do with it. MTAG performs an asymmetric multivariate test of all the phenotypes you input into it, which don't necessarily have the same properties as regular GWAS results. So I'm not sure whether the LAVA results would remain valid when run on MTAG output, especially if p-values for both phenotypes came from MTAG, and/or the one phenotype was involved in the MTAG analysis for the other.
Best, Christiaan
Hi,
Thank you so much for developing this amazing tool! I used genome-wide significant loci identified by MTAG to estimate local genetic correlations. However, the results in many regions are NA.
I summarized three NA scenarios in my results: 1) all estimates are NA, 2) only p-value is NA, 3) r2.lower, r2.upper, lo.lower and lo.upper are NA.
I found a related post explaining the third scenario. But what explains scenarios 1 and 2? Is there anything that can be done o solve the two problems?
I appreciate your help and time! Thank you!
Huifang