Post date: Jun 08, 2015 10:55:11 PM
I generated parameter estimates and diagnostics from the hdf5 output for the last few runs with:
perl ../scripts/runDiags.pl estpanc_n20*Chain0.hdf5 estpanc_n10demogGens*Chain0.hdf5
perl ../scripts/estParms.pl estpanc_n20*Chain0.hdf5 estpanc_n10demogGens*Chain0.hdf5
Note, some of the diagnostics are not great, I should look into this more.
I then used the calcCvrmsdPopQ*pl and plotCvrmsd*R scripts to generate summaries of the cvrmsd for population ancestry frequencies. Results for 10 vs. 20 SNPs are very similar and both beat 4 SNPs and the linkage model in structure. The comparison with structure was limited to the demographic simulations, but I am now running structure on the selection datasets, e.g.,
cd /local/scratch/
sleep 0
structure -i /labs/evolution/projects/popanc_sims/structure/infiles/genostructF_r9gen_strsel_gens50.txt -m /labs/evolution/projects/popanc_sims/structure//mainparams -e /labs/evolution/projects/popanc_sims/structure//extraparams -o out_struct_r9gen_strsel_gens50Chain0
scp out_struct_r9gen_strsel_gens50Chain0* /labs/evolution/projects/popanc_sims/structure/
cd /local/scratch/
After this I will just need to tweak the plots, add some numerical summaries, and then its on to the human data set(s).