FarmCPU

Association analyses with dominant markers were performed with the FarmCPU¹⁶⁷ method in R. For these analyses, markers were recoded to indicate the presence (0) and absence (2) of bands. We tested FarmCPU using no covariates and including matrices obtained from the three genetic structure analyses described in the previous section as such. In each case, a Q–Q plot of the -log₁₀(p) values of markers was generated, and the genomic inflation factor λ¹⁶⁸ was calculated. The average λ from analyses employing each covariate matrix was calculated and used to select the model that best controlled inflation. The Bonferroni correction with α = 0.05 was used to establish the significance threshold for associations, and the phenotypic variance explained by each marker was estimated for significant marker-trait associations using a linear model in R software.