Post date: Aug 22, 2014 7:8:55 PM
The runs from the BSLMM normal errors model were finishing and suggested a very hight PVE (~0.95). This seems really unlikely, and I suspect this is another failing of the linear approximation for a binary trait. So, I am trying the probit model instead, and with a slightly shorter run (hoping it will finish). Here is an example:
cd /home/A01963476/projects/timema_wgexperiment/gemma/
sleep 25
gemma -g geno_timemaC.txt -p pheno_samplesC.txt -a annotation.txt -bslmm 3 -n 1 -o gemoutProbit_timemaCrep5 -rpace 40 -w 1000000 -s 3000000