Why kinship is a random effect while population structure is a fixed effect in GWAS studies?

Question

[rimgubaev]

Hello everyone! The mixed linear models frequently applied for GWAS account for confounding factors like fixed (population structure) and random effects (kinship). I am confused with the following thing: The population structure information for example PCs are calculated based on genotype data and kinship matrix is also calculated based on genotype data. So basically both have one source but nevertheless considered as different types of confounding factors. I understand that population structure is definitely a fixed effect by drawing an analogy with herd in animal husbandry or a planting plot in horticulture. However it is still not clear why kinship should be considered as a random effect then.

[Researcher]

I have the same question.