Interpretation of the results of Wald test for association between lineages and phenotype

[neginmb]

Is the low p-value for the wald test of association between lineage and phenotype only an indicator of a positive association or also a negative association between them?

[johnlees]

Could be either, have a look at the value and sign of beta to infer this (positive sign, positive association with having the variant)

[neginmb]

Aha, thank you so much. You mean PCX, right? as I asked about the association of lineage rather than a variant.
I am asking this because I have a variant that without adding lineages is not significantly associated with the phenotype, but after considering population structure, it is very significantly associated with the phenotype. However, the most associated lineage for this variant (based on the wald test) is not associated with the phenotype. Do you have any explanation for this case?

[johnlees]

Sorry, I didn't read the lineage part of your message. We don't actually report the effect size/direction for the lineages, just the p-values – so yes could be in either direction, but you would have to do that regression yourself to determine which.

The variant in question may be present in multiple lineages, or there may be a more complex association than can be represented by a single top association. What I would recommend in cases like this would be to plot the presence of the variant on a phylogeny, along with the phenotype in question (e.g. using phandango). This usually helps determine whether the variant is restricted to a single lineage with high phenotype prevalence, or whether it is in multiple lineages