Ramdom model for LR finished: 1. when df1==1, the integrate lower boundary should be around .1, or it can't converge. 2. a little thing to do is to verify it now:
Verification results: with the result I obtained before, the prediction is not so satisfying. e.g. for nf=1, n=1000, q2=.05, h2=.4, with 1000 iterations, the empirical results are: 14.43, 60.1, 48.6 and 35.9. while the theory results are 15.1, 61.1, 50.6, and 39.5.
i will do a dense simulation test ... the empirical results almost hit the prediction precisely for single families!!!
... multiple family still needs to be adjusted. acturally it's no good, the estimates are too large. sth went wrong.
3 comments:
The integeration over the QTL distribution is a bit tricky. Algorithm needs to be optimized to speed up.
Ramdom model for LR finished:
1. when df1==1, the integrate lower boundary should be around .1, or it can't converge.
2. a little thing to do is to verify it now:
Verification results: with the result I obtained before, the prediction is not so satisfying. e.g. for nf=1, n=1000, q2=.05, h2=.4, with 1000 iterations, the empirical results are: 14.43, 60.1, 48.6 and 35.9. while the theory results are 15.1, 61.1, 50.6, and 39.5.
i will do a dense simulation test ... the empirical results almost hit the prediction precisely for single families!!!
... multiple family still needs to be adjusted. acturally it's no good, the estimates are too large. sth went wrong.
Post a Comment