Post date: Nov 22, 2013 10:36:35 PM
The poor behavior of the BSLMM model for survival in SLA on Ac could be caused by using a linear model for a binary trait (despite the good behavior of this model in other data sets). Consequently, I want to see whether using the probit model (-bslmm 3) produced a different and more reasonable estimate of pve. Thus, I am running the probit model on survival for each of the four experiments. I am running 5 chains each with a 500,000 iteration burn-in and 2 million MCMC steps. The jobs are in the long queue with ids 46749-46768 (the probit model takes much longer than the linear model).