1. plink 1.x --assoc has been obsolete for more than a decade. At minimum, it is standard to include top principal components (computed with e.g. --pca) as covariates in a logistic/Firth regression (
plink 1.x --logistic,
plink 2.0 --glm adds a fallback to Firth regression when logistic regression fails to converge in time). The top of the --glm documentation also links to several other programs implementing modern approaches.
2. You should also reread the documentation for the polygenic risk software package you are using.