Post date: Nov 25, 2013 9:12:59 PM
Using the probit model affected PVE and individual locus effects for survival. Most notably, the PVE and locus effects for SLA on Ac decreased substantially. With that said, I didn't run the MCMC long enough to get reliable parameter estimates. Thus, increased the MCMC estimation procedure to 5 million steps with a 1 million step burn-in and I am running 10 chains for each experiment (these are in the long queue with 280 hours). It would be interesting to do this for combined experiments too, but the trait is not binary after removing the treatment or population effect.