 |
 |
Dear Annchen, > I've done some experimenting with the --initrand option and have found that it seems to have no effect in the eddy_openmp version I have been running (with version 5.0.9 or 5.0.10), however I can use it to get consistent results with the eddy command in v5.0.9. But, I'd also like to use the --repol option which as I understand is only available in the 5.0.10 version. I'm not sure that I can use the _cuda version since I'm running everything on a cluster without cuda installed (and without admin access) - is there some way I can use both, or should I just not bother with it at this point? hmm, you say that eddy_openmp does not give you consistent results if you use --initrand? I have not come across that, but neither do I think I have necessarily tested that. Thinking about it I can see how running things multi-threaded might alter the sequence of random numbers. I will put it on my todo list to check that. If I were you I would not bother with --initrand, it will not have an impact on your final results. The --repol on the other hand will have an impact if you have data with lots of movement, and hence dropout. Jesper > > Thanks! > > Annchen >