Post date: Mar 26, 2015 9:14:29 PM
The first round of analyses for the paper generally look good. These involve 10 replicates of 20, 50, and 200 generations. Scale parameter estimates look good and make sense (smaller scale parameter for longer numbers of generations). Similarly, though not perfect, CVRMSD are not bad (they beat a null model with all loci equal), correlations between the true and estimated parameters are reasonable (~0.7), and the plots of true vs. estimated values look pleasing. There is still a slight tendency for the estimates to pull back towards the mean and 95% CI coverage is too low, but overall I think the results are good enough for this to be a useful method for me and others. The main results so far are in /labs/evolution/projects/popanc_sims/analysis/ and include two figures for the paper.