I have a genome-wide SNP dataset from RADseq containing 100 individuals from 4 groups. I converted the data so that it's in the form of a matrix (columns are SNPs, rows are individuals).
I'd like to use a constrained ordination (like a discriminant function) analysis in which I use the individuals from 2 of the 4 total groups to infer the axis that best separates these 2 groups across all SNPs. I then want to predict, along this axis, the position of the individuals of the other two groups.
I can easily convert the current SNP matrix into a numeric count matrix, such as by recoding, at each SNP, homozygous genotypes for allele A as 0, homozygous genotypes for the alternative allele B as 2, and heterozygous genotypes as 1.
My question is: what type of constrained ordination is appropriate for this type of data and goal? I'd like to do this in R.
I'd like to use a constrained ordination (like a discriminant function) analysis in which I use the individuals from 2 of the 4 total groups to infer the axis that best separates these 2 groups across all SNPs. I then want to predict, along this axis, the position of the individuals of the other two groups.
I can easily convert the current SNP matrix into a numeric count matrix, such as by recoding, at each SNP, homozygous genotypes for allele A as 0, homozygous genotypes for the alternative allele B as 2, and heterozygous genotypes as 1.
My question is: what type of constrained ordination is appropriate for this type of data and goal? I'd like to do this in R.