Post date: Jun 07, 2017 5:17:6 PM
ZG looked into how BSLMM (all inds) was going so far. The answer is well. a1 size has a reasonable pge and a high n_gemma (100+), whereas a1 position has a lower pge (~10%) but low n_gemma (there may just be one locus of major effect!).
Tonight we will work on scripts for plowing through the data to:
1. make a figure per group (all, each subgroup) that has each character along the x-axis, and plots pve and pge on the y (0-100%; bar plots with different colors), with n_gemma point estimated stated above the bars.
2. pip (posterior inclusion probability) for each SNP for each character for each Lycaeides group (these will have to be small plots so that all fit on a page for each group).
3. breeding values (variance in breeding values is additive genetic variance) - covariance matrix for each group for heat map and tests of modularity
pip = prob has an effect; multiply prob of having an effect x effect conditional on it having an effect = model average effect (pip and model ave effect are correlated)
overlaying pips for all size traits, all position; pairwise pip correlations to see if snps are same across traits